Metastable supergravity vacua with · 2007-10-05 · Metastable supergravity vacua with F and D...
Transcript of Metastable supergravity vacua with · 2007-10-05 · Metastable supergravity vacua with F and D...
Metastable supergravity vacua with
F and D supersymmetry breaking
Marta Gomez–Reino and Claudio A. Scrucca
Institut de Physique, Universite de Neuchatel,
Rue Breguet 1, CH-2000 Neuchatel, Switzerland
Abstract
We study the conditions under which a generic supergravity model involving
chiral and vector multiplets can admit viable metastable vacua with sponta-
neously broken supersymmetry and realistic cosmological constant. To do so,
we impose that on the vacuum the scalar potential and all its first derivatives
vanish, and derive a necessary condition for the matrix of its second derivatives
to be positive definite. We study then the constraints set by the combination
of the flatness condition needed for the tuning of the cosmological constant
and the stability condition that is necessary to avoid unstable modes. We find
that the existence of such a viable vacuum implies a condition involving the
curvature tensor for the scalar geometry and the charge and mass matrices for
the vector fields. Moreover, for given curvature, charges and masses satisfying
this constraint, the vector of F and D auxiliary fields defining the Goldstino
direction is constrained to lie within a certain domain. The effect of vector
multiplets relative to chiral multiplets is maximal when the masses of the vec-
tor fields are comparable to the gravitino mass. When the masses are instead
much larger or much smaller than the gravitino mass, the effect becomes small
and translates into a correction to the effective curvature. We finally apply
our results to some simple classes of examples, to illustrate their relevance.
1 Introduction
Recently substantial progress has been achieved in understanding how spontaneous super-
symmetry breaking could be realized in a phenomenologically and cosmologically viable
way in string inspired supergravity models. On one hand, from the microscopic string
point of view, the structure of the Kahler potential is well understood at leading order
in the weak coupling and low energy expansions [1, 2], and also the effects leading to
non-trivial superpotentials are now believed to be reasonably well understood [3, 4, 5],
although the identification of viable models still remains an open problem. Actually, it
has been suggested that perhaps one should not look for a unique candidate that would
be singled out for some reason, but rather for a statistical distribution in the landscape of
possible vacua [6]. On the other hand, from the macroscopic supergravity point of view,
much progress has been made in understanding the structure that the Kahler potential
and the superpotential need to have in order for the theory to admit a phenomenologically
acceptable vacuum. It has also been argued that the tuning of the cosmological constant
could be realized in a very economical and transparent way in a theory including two sec-
tors, one of them admitting an AdS supersymmetric vacuum and the other one breaking
supersymmetry and adding a positive contribution to the potential [7].
Spontaneous supersymmetry breaking can be triggered both by the F auxiliary fields
of chiral multiplets and the D auxiliary fields of vector multiplets. It is however believed
that the qualitative seed for the breaking must come from chiral multiplets, and that
vector multiplets may only affect the quantitative aspects of it. The reason for this is that
in standard situations the values of the D’s turn out to be proportional to the values of the
F ’s at any stationary point of the superpotential. This is true both in rigid [8] and local
supersymmetry [9], and relies essentially on the holomorphicity of the superpotential. The
only situation where this relation can be possibly avoided is in the presence of a genuine
field-independent Fayet-Iliopoulos term. However, although in global supersymmetry this
is a natural possibility, in local supersymmetry it is highly constrained due to the fact that
the usual Fayet-Iliopoulos term is not invariant in the presence of gravity [10]. In fact,
it turns out that even in the presence of a constant Fayet-Iliopoulos term associated to a
gauged R symmetry, the D’s are still proportional to the F ’s, as long as the superpotential
does not vanish [11], as is required in order to achieve a finite supersymmetry breaking
scale with vanishing cosmological constant. Nevertheless, despite this relation between
the F and D auxiliary fields, there are scenarios in which the D-terms may play a crucial
role besides the F -terms in producing a viable vacuum.
The situation in models where supersymmetry breaking is dominated by chiral multi-
plets is by now well understood. It is known that the form of the Kahler potential imposes
crucial restrictions on the possibility of getting a viable vacuum, that is, a metastable
stationary point with a very tiny cosmological constant. It was for example argued in
several ways that just the dilaton modulus could not lead to a viable situation [12], unless
subleading corrections to its Kahler potential become large [13, 14], and therefore the
perturbative control over the theory is compromised. Similarly, it was realized that the
compactification volume modulus could instead dominate the supersymmetry breaking,
2
but only if subleading corrections to its Kahler potential are taken into account [15, 16].
In two recent papers [17, 18], we have addressed in a more systematic way the general
question of whether it is possible to translate more directly onto the parameters of a given
theory the condition of flatness, related to the cancellation of the cosmological constant,
and the condition of stability, related to the stabilization of all the fluctuation modes. We
found that, in the simplest case of supergravity models involving only chiral multiplets,
it is possible to answer to this question in a remarkably sharp and simple way. It turns
out that there exists a strong necessary condition on the value that the curvature of the
Kahler geometry is allowed to take. Moreover, the Goldstino vector of F auxiliary fields
gets constrained not only on its length, but also on its direction. More precisely, the
form of the Riemann tensor must present valleys where the curvature takes a value that is
below a certain threshold, and the Goldstino direction cannot be too far away from those
directions, how far this can be being fixed by the value of the curvature scalars.
The results of [17, 18] provide a criterium that allows to discriminate to some extent
between promising and non-promising models by knowing the approximate form of their
Kahler potential only, independently of the form of the superpotential. Indeed, they were
shown to have very useful and general implications on the possibility of realizing in a
viable way supersymmetry breaking when it is dominated by the F auxiliary fields. These
implications are particularly striking for string models, where the moduli sector is identi-
fied with the hidden sector. The main reason for this is that in these models the Kahler
geometry has constant curvature at leading order in the coupling and derivative expan-
sions, so that the constraints translate very directly into restrictions on the parameters
of the Lagrangian. For instance, the lower bound found for the curvature was used to
explain in a more robust way why the dilaton cannot dominate supersymmetry breaking,
whereas the volume modulus can dominate it but in a way that is very sensitive to small
subleading corrections to the Kahler potential. A similar strategy has also been used to
explore the statistics of supersymmetry breaking vacua in certain classes of string mod-
els [19]. Finally, there have also been studies adopting a complementary viewpoint and
developing more efficient tools to study algebraically the vacua allowed by a given theory
with fixed Kahler potential and superpotential [20].
The situation in models where supersymmetry breaking is significantly affected not
only by chiral multiplets, but also by vector multiplets, is more complex and less under-
stood. The fact that the D-term contribution to the scalar potential is positive definite,
(unlike the F -term contribution, which has an indefinite sign), suggests that they may play
an important role in the stabilization of the scalar fields triggering spontaneous supersym-
metry breaking. Note that this distinction is essentially implied by Lorentz invariance,
which allows scalar fields to have a generic potential leading to non-vanishing VEVs, but
requires vector fields to have a restricted potential leading to vanishing VEVs. As a result,
the squared masses of all the charged fields tend to be increased by the gauging. Actually
there are two types of effects. The first is that Goldstone fields corresponding to global
symmetries of the ungauged theory get absorbed by gauge fields through the Higgs mech-
anism when the coupling is switched on, and thereby obtain a positive physical squared
mass. The second is that the non-Goldstone fields also receive a positive correction to their
3
squared masses due to the gauging, which can therefore help in stabilizing them. These
possibilities have been recently explored for instance in [8, 21, 22, 23, 24, 25, 26, 27], but
a more systematic and quantitative understanding of the features of this more general
situation is still missing. The aim of this work is to try to fill this gap by performing
a general study from which one could extract information that can be directly used in
practice to build viable models.
The main goal of this paper is thus to generalize the study performed in [17, 18] to
the more general class of supergravity theories involving not only chiral multiplets but
also vector multiplets that gauge isometries of the Kahler geometry. This can actually be
done by following the same strategy as in [17, 18]. Conceptually, the main novelties lie in
the fact that besides the Kahler potential and the superpotential, one must also specify
the Killing vectors defining the isometries that are gauged and the gauge kinetic function
defining the couplings. Technically the task is substantially more complicated due to the
fact that the Goldstino vector now involves not only F auxiliary fields but also D auxiliary
fields, and also that the scalar potential has a more complicated structure. Nevertheless,
it is possible to find the exact generalization of the flatness and stability conditions of
[17], but their implications can be worked out in detail only in certain specific regimes,
due to the increased algebraic complication of the problem. In general, the effect of
vector multiplets alleviates the necessary conditions for flatness and stability, compared
to the case involving only chiral multiplets. However, the associated gauge symmetries also
restrict the variety of models, as the superpotential should be gauge invariant up to Kahler
transformations. We will actually show that in a variety of situations, the net qualitative
effect of vector multiplets is to reduce the effective curvature felt by the chiral multiplets.
In certain situations, this can be more sharply interpreted as coming from a correction
to the effective Kahler potential induced by the presence of vector multiplets. As already
mentioned, corrections to the Kahler potential can be crucial for the existence of a viable
vacuum in certain situations. In this respect, there is a very interesting distinction between
corrections induced by extra chiral multiplets and those induced by vector multiplets: the
former can either increase or decrease the curvature, since they have an indefinite sign,
whereas the latter (as we will show in the body of the paper) always decrease it.
A final comment is in order regarding the issue of implementing the idea of getting a
metastable Minkowski minimum through an uplifting sector that breaks supersymmetry
in a soft way. It is clear that such a sector will have to contain some light degrees of
freedom, providing also some non-vanishing F and/or D auxiliary field. Models realizing
an F -term uplifting are easy to construct. A basic precursor of such models was first
constructed in [28] and then further exploited, for instance, in [29]. More recently, a
variety of other examples have been constructed, where the extra chiral multiplets have
an O’ Raifeartaigh like dynamics, which is either genuinely postulated from the beginning
[30, 31] or effectively derived from the dual description of a strongly coupled theory [32]
admitting a metastable supersymmetry breaking vacuum as in [33]. Actually, a very
simple and general class of such models can be constructed by using as uplifting sector
any kind of sector breaking supersymmetry at a scale much lower than the Planck scale
[17]. Models realizing a D-term uplifting, on the other hand, are difficult to achieve. The
4
natural idea of relying on some Fayet-Iliopoulos term [34] does not work, due to the already
mentioned fact that such terms must generically be field-dependent in supergravity, so that
the induced D is actually proportional to the available charged F ’s. It is then clear that
there is an obstruction in getting D much bigger than the F ’s. Most importantly, if the
only charged chiral multiplet in the model is the one of the would-be supersymmetric
sector (which is supposed to have vanishing F ) then also D must vanish, implying that a
vector multiplet cannot act alone as an uplifting sector [35, 36]. This difference between
F -term and D-term uplifting is once again due to the basic fact that chiral multiplets can
dominate supersymmetry breaking whereas vector multiplets cannot.
The paper is organized as follows: In section 2, we present a brief summary of the
relevant features of gauge invariant supergravity models. In section 3 we discuss the
general properties of the vacuum and the interplay between the spontaneous breaking of
supersymmetry and gauge symmetries. In section 4 we study various types of relations
that exist between D and F auxiliary fields. In section 5 we derive a necessary condition for
stability, and combine it with the flatness condition to define a set of general constraints
that are necessary for the existence of a viable vacuum. In section 6 we elaborate on
the possible strategies that can be used to derive more concrete implications from these
constraints, and in particular on the different ways in which the relation between D and
F auxiliary fields can be taken into account. In sections 7, 8 and 9 we then pursue three
different approaches to this problem, which are based respectively on a dynamical relation,
a kinematical relation and a kinematical bound between the D and the F auxiliary fields.
In section 10 we illustrate the relevance of our general results with some examples of string
inspired models. In section 11, we conclude with a qualitative summary of our results.
2 Gauge invariant supergravity models
In this section we will briefly review the main features of general supergravity models with
minimal supersymmetry in four dimensions [37, 38], emphasizing those particular aspects
that will be relevant for our analysis. We will use the notation of [39] and set MP = 1.
Consider first a supergravity theory with n chiral multiplets Φi. The two-derivative
Lagrangian is specified by a single real Kahler function G(Φk,Φk†) 1. The Kahler geometry
of the manifold spanned by the complex scalar fields is determined by the metric gij = Gij,
which can be used to raise and lower chiral indices 2. It can happen that this theory
has a group of some number m of global symmetries, compatibly with supersymmetry.
These are generated by holomophic Killing vectors Xia(Φ
k), in the sense that a generic
symmetry transformation with infinitesimal real constant parameters λa is implemented
by the operator δ = λa(Xia∂i+X
ia∂i). The chiral superfields transform as δΦi = λaXi
a(Φk).
It is then clear that the condition for these transformations to represent an invariance of
1The function G is related to the more commonly used Kahler potential K and superpotential W by
the equation G = K + log|W |2. This decomposition is however ambiguous, due to the Kahler symmetry
transforming K → K +F + F and W → e−FW , and leaving G invariant. As already mentioned, W cannot
vanish at a non-supersymmetric Minkowski minimum, and the function G is therefore well defined.2Subscripts on scalar quantities denote ordinary derivatives with respect to the fields.
5
the theory is that the function G should be invariant: δG = 0. This implies:
XiaGi +X i
aGi = 0 . (2.1)
This equation is valid at any point of the scalar manifold and one can therefore derive
additional conditions by taking derivatives of it. The corresponding information is most
conveniently extracted by using covariant derivatives ∇i, in terms of which holomorphicity
of the vectors Xia implies ∇iX
ja = 0 or ∇iXaj = 0. Taking one derivative of eq. (2.1), one
finds that:
Xai +Xka∇iGk +Gk∇iX
ka = 0 . (2.2)
Taking two derivatives, and using the fact that the metric is covariantly constant, one
finds instead:
∇iXaj + ∇jXai = 0 . (2.3)
Using the definition of the Riemann tensor and the holomorphicity of the Killing vectors
we know that ∇i∇jXap = RijpqXqa. With the help of this relation, and taking a suitable
combination of four derivatives of (2.1), one also finds:
(
Xma ∇m+X n
a ∇n
)
Rijpq +Rrjpq∇iXra +Rispq∇jX
sa +Rijtq∇pX
ta +Rijpu∇qX
ua = 0 . (2.4)
The conditions (2.3) and (2.4) can be rephrased in terms of Lie derivatives as LXa gij = 0
and LXa Rijpq = 0, and show that each symmetry is associated to an isometry of the scalar
manifold 3.
Consider now the possibility of gauging such isometries with the introduction of vector
multiplets. The corresponding supergravity theory will then include n chiral multiplets
Φi and m vector multiplets V a. Its two-derivative Lagrangian is specified by a real Kahler
function G(Φk,Φk†, V a), determining in particular the scalar geometry, m holomorphic
Killing vectors Xia(Φ
k), generating the isometries that are gauged, and an m by m matrix
of holomorphic gauge kinetic functions Hab(Φk), defining the gauge couplings. There
exists a general and systematic way of promoting the globally invariant action of the chiral
multiplet theory to a locally invariant Lagrangian involving also the vector multiplets. This
can be done in superfields by generalizing the real constant transformation parameters λa
to chiral superfield parameters Λa, and asking the transformation operator δ to act also
on the vector fields, as δ = ΛaXia∂i + ΛaX i
a∂i − i(
Λa − Λa)∂a. The transformations of
the chiral and vector multiplets read then δΦi = ΛaXia(Φ
k) and δV a = −i(Λa − Λa).
The minimal coupling between chiral and vector multiplets turn ordinary derivatives into
covariant derivatives, and induces a new contribution to the scalar potential coming from
the vector auxiliary fields Da, in addition to the standard one coming from the chiral
auxiliary fields F i. The condition for the action to be invariant under the just mentioned
3Notice that for constant curvature Kahler manifolds, for which ∇kRijpq = 0, the conditions (2.4)
represent linear constraints on the derivatives of the Killing vectors.
6
local transformations is clearly that G must be invariant: δG = 0 4. This implies now the
conditions:
Ga = −iXia Gi = iX i
a Gi . (2.5)
In addition, the gauge kinetic function Hab must have an appropriate behavior under
gauge transformations, in such a way as to cancel possible gauge anomalies Qabc and to
lead to a consistent quantum effective action. More precisely, ReHab must be invariant,
whereas ImHab must have a variation that exactly matches the coefficient of Qabc5. In
general, such anomalies have a quite involved structure and can depend on the scalar fields.
Their general form has been studied in [41] in the limit of rigid supersymmetry, and more
recently in [42] in the context of local supersymmetry. The functional form that Hab is
allowed to take is then strongly constrained and linked to the form of the anomaly Qabc.
More precisely, the anomaly cancellation condition δHbc = i λaQabc implies the functional
relations
Xia∇iHbc = iQabc . (2.6)
Now, since the quantities Xia and Hbc are both holomorphic, these equations are subject
to a strong integrability condition. Indeed, by taking first derivatives of the relations
(2.6), one finds that the real function Qabc must actually be constant. The functional
form of the gauge kinetic function is then essentially fixed by the transformation rules
of the scalar fields, in such a way that its variation is constant. This means that the
possibility of canceling anomalies through local Wess-Zumino terms, which always exists
in general, is quite strongly constrained by supersymmetry. In the low-energy effective
theories underlying string models, for instance, it is realized in the special form of the
Green-Schwarz mechanism [43]. In that case, some fields that were neutral at tree-level
and present in Hab acquire a non-trivial transformation law due to one-loop corrections
to the Kahler potential. In addition, there can also be one-loop threshold corrections to
the gauge kinetic function that depend on charged fields. These two effects account for
anomaly cancellation in all the known situations.
The expectation values of the real and imaginary parts of Hab define respectively the
inverse couplings and the θ-angles for the vector fields. The former defines also a metric
for the gauge fields, which can be used to raise and lower vector indices:
hab = ReHab . (2.7)
The expectation values of the first and second derivatives of Hab define additional relevant
parameters of the theory. As a consequence of the holomophicity of Hab, it is possible to
4This means that the Kahler potential K and the superpotential W must be invariant only up to a
local Kahler transformation, associated with a gauging of the R symmetry.5In general, there can also be generalized Chern-Simons terms, which can be added to the Lagrangian
compatibly with supersymmetry. These can contribute to the anomalous variation of the action and allow
therefore to generalize a bit the way in which anomaly cancellation is realized. See [40] for discussions of
this point. Here we shall simply interpret Qabc as the residual quantum anomaly that is left after having
considered possible generalized Chern Simons terms.
7
parametrize these through the following two quantities:
habi = ∇iReHab , habij = ∇i∇jReHab . (2.8)
The condition (2.6) for anomaly cancellation implies then the relation:
Xiahbci =
i
2Qabc . (2.9)
The theory is most conveniently formulated by using the superconformal formalism
[44], with a chiral compensator multiplet Φ. The Lagrangian reads then simply 6:
L =
∫
d4θ[
−3 exp{
− 1
3G(Φk,Φk†, V a)
}]
Φ†Φ +
(∫
d2θΦ3 + h.c.
)
+
(∫
d2θ1
4Hab(Φ
k)W aαW bα + h.c.
)
. (2.10)
The kinetic Lagrangians for the scalar and the gauge fields involve respectively the covari-
ant derivative DµΦi = ∂µΦi −Xia(Φ
k)Aaµ and the field strengths F a
µν , and read:
Lkin = − gij DµφiD∗µφ
j − 1
4hab F
µνaF bµν . (2.11)
The scalar potential can be computed by integrating out the auxiliary fields. To do so,
it is convenient to gauge fix the redundant superconformal symmetries by setting the
scalar component of the compensator to eG/6 and the fermion component to 0, and to
parametrize its non-trivial auxiliary component as eG/6F . The auxiliary field Lagrangian
is then easily computed, and the corresponding equations fix F = eG/2(1−1/3GkGk) and
F i = − eG/2 gijGj , Da = −habGb . (2.12)
Substituting these expressions back into the auxiliary field Lagrangian, the scalar potential
is found to be:
V = eG(
GkGk − 3)
+1
2GaGa . (2.13)
Using the relation (2.5) and the expression m3/2 = eG/2 for the gravitino mass, one can
finally rewrite this in the standard way as
V = −3m23/2 + gij FiFj +
1
2habDaDb . (2.14)
where
Fi = −m3/2Gi , (2.15)
Da = iXia Gi = −iX i
aGi . (2.16)
6The D and F densities contain now, besides the terms appearing in rigid supersymmetry, also terms
depending on the graviton and the gravitino, but these will not be relevant for us.
8
The relation (2.16) shows that the Da can be geometrically identified with the Killing
potentials. Indeed, taking one derivative one finds that the Killing vectors can be written
in the following way, which automatically solves the Killing equation (2.3):
Xai = i∇iDa , Xai = −i∇iDa . (2.17)
Taking two derivatives of the relation (2.16), one obtains instead a two-index tensor char-
acterizing the local ”charges” of the chiral multiplets
qaij =i
2
(
∇iXaj −∇jXai
)
= ∇i∇jDa . (2.18)
Notice finally that by taking derivatives of (2.5), one also deduces that Xai = −iGai,
Xai = iGai and qaij = −Gaij.
3 Vacuum and spontaneous symmetry breaking
The vacuum of the theory is associated to a stationary point of the scalar potential (2.13).
The scalar fields take in general non-vanishing vacuum expectation values, and local super-
symmetry and the gauge symmetries can thus be spontaneously broken. The cosmological
constant scale is identified with the vacuum energy, and we will assume that it is adjusted,
through a tuning of parameters in the effective Lagrangian, to its very small measured
value, which is negligible with respect to the gravitino mass scale eG/2. We shall therefore
impose from the beginning that on the vacuum V = 0, implying the flatness condition:
−3 +GiGi +1
2e−GGaGa = 0 . (3.1)
The stationarity conditions correspond to requiring that ∇iV = 0, and they are given by:
Gi +Gk∇iGk + e−G[
Ga(
∇i −1
2Gi
)
Ga +1
2habiG
aGb]
= 0 . (3.2)
The 2n-dimensional mass matrix for small fluctuations of the scalar fields around the
vacuum has two different n-dimensional blocks, which can be computed as m2ij = ∇i∇jV
and m2ij = ∇i∇jV . Using the flatness and stationarity conditions, one finds, after a
straightforward computation [45, 46]:
m2ij = eG
[
gij −RijpqGpGq + ∇iGk∇jG
k]
+[
− 1
2
(
gij −GiGj
)
GaGa +(
G(ihabj) + hcdhacihbdj
)
GaGb (3.3)
− 2GaG(i∇j)Ga − 2Gahbchab(i∇j)Gc + hab∇iGa∇jGb +Ga∇i∇jGa
]
,
m2ij = eG
[
2∇(iGj) +Gk∇(i∇j)Gk
]
+[
− 1
2
(
∇(iGj) −GiGj
)
GaGa +(
G(ihabj) + hcdhacihbdj −1
2habij
)
GaGb
− 2GaG(i∇j)Ga − 2Gahbchab(i∇j)Gc + hab ∇iGa∇jGb
]
. (3.4)
9
The m-dimensional mass matrix for the vector fields can instead be read off from the
kinetic term of the scalar fields, and has the form:
M2ab = 2 gijX
iaX
jb = 2 gij ∇iGa∇jGb . (3.5)
In general the theory displays a spontaneous breakdown of both supersymmetry and
gauge symmetries. These two breakings can happen at independent scales and along
different directions in field space, and this is the main reason for the increased complexity
of this analysis. Supersymmetry breaking is realized through a fermionic version of the
Higgs mechanism, in which the gravitino field absorbs a linear combination of the fermion
fields ψi and ψa of the chiral and vector multiplets, the would-be Goldstino fermion. The
relevant combination can be read off from the mixing between the gravitino and the chiral
and vector multiplet fermions, and is given by:
η =i√3eG/2Giψ
i +1√6Gaψ
a . (3.6)
The Goldstino is thus associated with the vector ηξ = (ieG/2Gi/√
3, Ga/√
6) in the space
of chiral and vector multiplet spinors ψξ = (ψi, ψa): η = ηξψξ. The norm squared of this
vector is given by ηξηξ = eGGiGi/3+GaGa/6 = m2
3/2, due to the flatness condition (3.1),
and the scale of supersymmetry breaking can therefore be associated with the gravitino
mass m3/2. On the other hand, gauge symmetry breaking is realized through an ordinary
bosonic Higgs mechanism, in which the gauge fields Aaµ absorb linear combinations of
the scalar fields φi of the chiral multiplets, the would-be Goldstone bosons. The relevant
combinations can be read off from the mixing between the gauge fields and the chiral
multiplet scalar fields, and are given by:
σa = Xai φi +Xaiφ
i . (3.7)
The Goldstone bosons are thus associated to the vectors σaα = (Xai,Xai) in the space
of chiral multiplet scalars φα = (φi, φi): σa = σaαφα. The scalar product between two
of these vectors is given by σaασαb = 2Xi
aXbi = M2ab, and the scales of gauge symmetry
breaking are therefore controlled by the gauge field mass matrix Mab.
In general, supersymmetry and gauge symmetry breaking occur in an entangled way
because of two reasons. The first is that the directions of the breakings are in general not
orthogonal, and the second is that the breaking scales are not necessarily well separated.
This situation simplifies (and the two types of breakings disentangle) only whenever the
Goldstino and the Goldstone directions are nearly orthogonal or the gravitino and the
gauge field masses are hierarchically different. In the first case, the two breaking disentan-
gle just because they involve different sets of fields, whereas in the second they decouple
because of the very large mass difference of the relevant fields.
4 Chiral versus vector auxiliary fields
The chiral and vector auxiliary fields are given by Fi = −eG/2Gi and Da = −Ga. These
relations are enforced by the equations of motion of the auxiliary fields, and are therefore
10
true at any point of the scalar field space, and not only at the stationary points of the
potential. These two kinds of auxiliary fields are however not completely independent from
each other. There is a first relation between the Fi and the Da, which is of ”kinematical”
nature, in the sense that it is satisfied as a functional relation valid at any point of the
scalar field space. This first relation comes from (2.5) and is a consequence of the gauge
invariance of G. It reads:
Da = −i Xia
m3/2Fi = i
X ia
m3/2Fi . (4.1)
This relation shows that the Da are actually linear combinations of the Fi, with coefficients
of order O(Maa/m3/2). One can then derive a simple bound on the sizes that the Da can
have relative to the Fi. Indeed, using the inequality |aibi| ≤√
aiai
√
bjbj, one deduces
from the relation (4.1) that:
|Da| ≤1√2
Maa
m3/2
√
F iFi . (4.2)
There is also a second relation between the Fi and the Da, which is instead of ”dynam-
ical” nature, in the sense that it is valid only at the stationary points of the potential. It
comes from considering a suitable linear combination of the stationarity conditions (3.2)
along the direction Xia, that is, imposing stationarity with respect to those particular
field variations that correspond to complexified gauge transformations: Xia∇iV = 0. This
corresponds essentially to extract the information concerning stationarity of those scalar
fields that are absorbed by the gauge field in the Higgs mechanism. To derive this relation,
one starts by contracting (3.2) with Xia. The result can be simplified by using eq. (2.2).
Using also the relations (2.9), (2.16) and (3.5) one gets:
−∇iXaj FiF j − i
(
F iFi −m23/2
)
Da − i
2M2
abDb +
i
2QabcD
bDc = 0 . (4.3)
The real part of this equation is identically satisfied at any point. Indeed, the real part of
the first term vanishes due to the Killing condition (2.3), whereas that of the last three
terms vanishes trivially due to the fact that the quantities Da, M2ab and Qabc are real. This
just reflects the fact that the potential V is invariant under gauge transformations at any
point: Re(Xia∇iV ) = δV = 0. The imaginary part of this equation is instead non-trivial,
as V is not invariant under imaginary gauge transformations. Using the definition (2.18),
we get the following quadratic relation between the Da and the Fi [9, 47] (see also [48]):
qaijFiF j − 1
2
[
M2ab + 2
(
F iFi −m23/2
)
hab
]
Db +1
2QabcD
bDc = 0 . (4.4)
It is worth discussing what happens in the limit of rigid supersymmetry. In order to
do so, it is important to note that all the formulae that have been derived up to now in
this section are actually valid for any value of the cosmological constant, and not only
for the case where it vanishes. This means that the sizes of the auxiliary fields Fi and
Da are not necessarily limited by the flatness condition. We can then consider the limit
of rigid supersymmetry, m3/2 → 0 and MP → ∞, while keeping all the other quantities
11
fixed. In that limit, the kinematical constraint (4.2) becomes trivial. This reflects the fact
that in the rigid limit, the functional forms of the Fi and the Da simplify and become
independent of each other. On the other hand, the dynamical relation (4.4) persists in the
rigid limit, although it gets substantially simplified. Indeed, the second and third terms
in the relation (4.4) can be neglected, since by dimensional analysis they are suppressed
by powers of MP . Therefore we are left with: qaijFiF j − 1/2M2
abDb + 1/2QabcD
bDc = 0,
and we can conclude that the kinematical bound (4.2) is of gravitational origin, whereas
the dynamical relation (4.4) is of non-gravitational origin. In the light of this observation,
it should be emphasized that the obstruction in getting a non-zero D in situations with
vanishing F should be interpreted as a constraint coming from the dynamical relation
(4.4) rather than from the kinematical relation (4.1), since it is due to the holomorphicity
of the superpotential and not to any particular feature associated to gravity 7. It is clear
that this obstruction also translates into a natural tendency for D auxiliary fields to be
smaller than F auxiliary fields 8.
We are going to discuss now two situations in which the relation between the Fi
and the Da simplifies. They correspond to the two different limits where the breaking
of supersymmetry and gauge symmetries happen at well separated scales and therefore
disentangle.
4.1 Heavy vector limit
A first limit in which the situation substantially simplifies is when the gauge bosons masses
are much bigger than the gravitino mass: Mab ≫ m3/2. In this limit (and assuming that
all the other quantities in the problem remain finite) one finds that the kinematical bound
(4.2) becomes irrelevant, whereas the dynamical relation (4.4) implies that the Da are
small and given by
Da ≃ 2M−2abqbijFiF j . (4.5)
More precisely, the flatness condition implies that Fi ∼ O(m3/2), and the expression (4.5)
has thus the form of a sum of terms that are O(m3/2/Mab), with O(qaij) coefficients. The
simplification of the relation (4.4) in this limit is due to the fact that gauge symmetries are
broken at a much higher scale than supersymmetry. It is then a good approximation to
integrate out the vector multiplets in a supersymmetric way at the scale defined by Mab,
and consider supersymmetry breaking within a low-energy effective supergravity theory
involving only chiral multiplets. The Fi represent then the dominant supersymmetry
breaking effects that are directly induced by chiral multiplets, whereas the Da can now
be understood as encoding the subleading corrections to supersymmetry breaking effects
that are indirectly induced by virtual heavy vector fields [8].
Note also that due to the kinematical relation (4.1), the linear combination of chiral
auxiliary fields XiaFi (which corresponds to the vector auxiliary fields Da) must be small.
7Nevertheless, it is curious that in the presence of an anomaly cancelled by a Wess-Zumino term, eq.
(4.4) does not seem to display automatically this obstruction and one apparently needs to invoke (4.1).8See for instance [49] for a discussion on the possibility of achieving D larger than F in the simplest
class of theories with global supersymmetry and linearly realized gauge symmetries.
12
One gets then the m restrictions XiaFi ≃ 0 as a consequence of the fact that, in the limit
considered here, m of the n chiral multiplets are absorbed through a supersymmetric Higgs
mechanism, and only n −m gauge-invariant combinations of chiral multiplets remain as
light fields in the low energy effective action that is used to describe supersymmetry
breaking effects. Note that this indirectly implies that the vector fields can be much
heavier than the gravitino (without any other parameter blowing up) only if the number
of chiral multiplets is higher than the number of vector multiplets: n > m.
It is possible to verify that one correctly reproduces (4.5) by integrating out explicitly
the vector fields. This can be done by solving the vector multiplet equations of motion
directly in superfields. In the limit we are considering terms involving supercovariant
derivatives can be neglected, since on dimensional grounds they would give subleading
contributions. This means that one can neglect the gauge kinetic term in (2.10) and the
equations of motion of the vector superfields are simply:
Ga(Φi,Φ†i , Va) exp
{
− 1
3G(Φi,Φ
†i , Va)
}
Φ†Φ ≃ 0 . (4.6)
These represent algebraic equations involving the vector superfields Va, the chiral super-
fields Φi, and the compensator superfield Φ. To solve them and integrate out the vector
superfields, one needs to choose a gauge for the superfield gauge transformations. A con-
venient choice is the super-unitary gauge, where the would-be Goldstone superfields are
set to zero. At the linearized level, this means setting Ga ≃ 0 (the ≃ sign being related
to the fact that this is possible only in the exactly supersymmetric limit). One can then
solve eq. (4.6) and express the V a in terms of the physical Φi orthogonal to Xia and Φ.
Substituting the results back into the Lagrangian, one would then find the whole effective
Lagrangian for the light chiral multiplets. Except for very special models, this is however
difficult to carry out explicitly in practice, due to the fact that arbitrary powers of V a
appear in the equation. As a consequence, it is in general not possible to solve (4.6)
exactly at the superfield level to determine the complete effective Lagrangian in closed
form. One can however solve it approximately by assuming that the superfields V a are
small, and expanding it in powers of V a. At first order in the expansion, one finds that
this is consistent if Ga ≃ 0, as is the case in the super-unitary gauge we are using. At
second order, the equation for V a is linear and can be solved to determine the leading
order effect of the vector multiplet. In doing so one finds the following expression:
[
Gab −1
3GaGb
]
V b +Ga ≃ 0 . (4.7)
Taking into account that the quantity Gab−GaGb/3 has a vacuum expectation value given
by M2ab/2, which is by assumption large, one finally finds:
V a ≃ −2M−2abGb . (4.8)
In this equation, V a and Gb are superfields, whereas M−2ab are numbers. Taking finally
the D component of this equation, and recalling that Gaij = −qaij, one recovers eq. (4.5).
One can also use the result (4.8) to compute the leading order correction that is induced
by the heavy vector multiplets in the effective action for the light chiral multiplets. It is
13
given by ∆G ≃ −M−2abGaGb9. Notice finally that the F component of the gauge fixing
condition Ga ≃ 0 reproduces instead the constraint XiaFi ≃ 0, since Gai = iXai.
4.2 Light vector limit
Another limit in which the situation also simplifies is when the gauge bosons masses are
much smaller than the gravitino mass: Mab ≪ m3/2. In this limit (and assuming that all
the other quantities in the problem remain finite) the kinematical bound (4.2) becomes
very stringent and implies that the Da are small. By the flatness condition we get then
F iFi ≃ 3m23/2, and the dynamical relation (4.4) simplifies to
Da ≃ 1
2m−2
3/2 qaijFiF j . (4.9)
Note that since Fi ∼ O(m3/2), this expression has the form of a sum of terms that are
O(1), with O(qaij) coefficients. But from the kinematical bound we know that theDa must
actually be small, more precisely O(Mab). This implies that there must be a cancellation
of the leading order contribution of the various terms: qaijFiF j ≃ 0. This shows that
the vector multiplets can be very light only if the charged chiral multiplets contribute to
supersymmetry breaking in a suitably aligned way. It also indirectly implies that, unless
the charges qaij become somehow degenerate, the vector fields can be much lighter than
the gravitino only if the number of chiral multiplets is higher than the number of vector
multiplets: n > m.
The fact that the relation between the Da and the Fi simplifies in this case can be
partly understood from the fact that in this limit supersymmetry is broken at a scale
much higher than the scale of gauge symmetry breaking. The Da must then be small, as a
consequence of the fact that supersymmetry breaking is dominated by the Fi of the chiral
multiplets. On the other hand, in order to approximately preserve gauge invariance, the
vacuum expectations values of the latter cannot be arbitrary, but must rather satisfy m
approximate relations, which are identified with the conditions qaijFiF j ≃ 0.
5 Constraints from flatness and stability
We want now to study the constraints that can be put on the theory by imposing the flat-
ness condition, guaranteeing the vanishing of the cosmological constant, and the stability
condition, enforcing the positivity of the scalar mass matrix. In order to do so, it is useful
to introduce the 2n-dimensional vector of scalar fields φα =(
φi, φi)
. Using this notation,
the complete 2n-dimensional mass matrix for the scalar fields can be written in terms of
the blocks (3.3) and (3.4) as
m2αβ =
(
m2ij m2
ij
m2ij m2
ij
)
. (5.1)
9The correction DaDa/2 to the scalar potential is not included in this result, and corresponds to a
subleading effect coming from the vector field kinetic terms.
14
In order to study the restrictions imposed by the requirement that the physical squared
mass of the scalar fields are all positive, it is necessary to take appropriately into account
the spontaneous breaking of gauge symmetries. In this processm of the scalars, the would-
be Goldstone bosons, are absorbed by the gauge fields and turned into their longitudinal
components. In the unitary gauge, these modes are disentangled from the remaining
physical scalar fields, and their m-dimensional mass matrix coincides with the mass matrix
M2ab of the gauge fields. Since this matrix is by construction positive definite, we do not
need to worry about these directions for the stability of the physical scalar fields. What
we have instead to impose is that the mass matrix m2αβ of the remaining 2n−m physical
scalar fields is positive definite. This complicates the analysis with respect to the case
where vector multiplets are absent, because the matrix m2αβ is defined by a complicated
real projection of the Hermitian matrix m2αβ onto the subspace orthogonal to the would-be
Goldstone directions Xαa = (Xi
a,Xia).
Fortunately, there exist a simple way to avoid the extra complications introduced
by the gauge fixing procedure. It relies on the observation that the would-be Goldstone
modes correspond, before the gauge fixing, to flat directions of the unphysical mass matrix
m2αβ, and get their physical mass through their kinetic mixing with the gauge bosons 10.
This means that positivity of the physical mass matrix m2αβ implies semi-positivity of
the unphysical mass matrix m2αβ. This condition is much simpler to study, and we will
therefore take it as our starting point.
To extract some useful information from the condition that the mass matrix (5.1)
should be semi-positive definite, we can now use the same strategy as in [17]. Instead
of requiring that m2αβ Z
αZ β ≥ 0 for any direction Zα, we can look only along some
special directions, and see if we can obtain simple necessary conditions. Note first that
combining the conditions coming from the two independent directions Zα1 = (zi, z i) and
Zα2 = (izi,−iz i) defined by any complex vector zi, one deduces that m2
ij zizj ≥ 0. This
means that if m2αβ
is semi-positive definite, then the principal block m2ij must necessarily
be semi-positive definite as well. Our strategy will then be, as in [17], to look for a suitable
vector zi leading to a simple constraint on the parameters of the theory.
In the case of theories with both supersymmetry and gauge symmetries, there exist
two types of special complex directions zi one could look at. The first is the direction
Gi, which is associated with the Goldstino direction in the subspace of chiral multiplet
fermions. The second is identified with the directions Xia, which are instead associated
with the Goldstone directions in the space of chiral multiplet scalars. Consider first
the direction Gi. Starting from (3.3) and using the fact that Gi∇iGa = Ga, the flatness
condition (3.1), and also the stationarity condition (3.2) to simplify terms involving ∇iGj ,
10It is straightforward to verify that the Goldstone directions Xαa correspond indeed to flat directions of
the mass matrix m2αβ = ∇α∇βV . To see this note that, in the notation we are using in this section, the
operator generating real gauge transformations is just δ = λaXαa∇α. The condition of gauge invariance
of the potential V , namely δV = 0, implies thus that Xαa∇αV = 0. Since this relation is valid at any
point, one can take another derivative and deduce the new condition ∇βXαa∇αV +Xα
a∇β∇αV = 0. Now,
evaluating this expression at a stationary point, where ∇αV = 0, one finds m2αβX
βa = 0, implying in
particular that m2αβXα
aX βa = 0.
15
one finds, after a long but straightforward computation:
m2ijG
iGj = eG[
6 −Rijpq GiGjGpGq
]
+[
−2GaGa + hcdhacihbdjGiGjGaGb
]
(5.2)
+ e−G[
M2abG
aGb− 3
4QabcG
aGbGc− 1
2
(
GaGa
)2+
1
4h i
ab hcdiGaGbGcGd
]
.
The condition m2ijG
iGj ≥ 0 is then the generalization of the condition derived in [17] for
theories involving only chiral multiplets. We believe that also in this case this condition
captures the most significant restrictions associated to stability, insofar it comes from the
scalar and pseudoscalar partners of the Goldstino, which are generically expected to be
the lightest and most dangerous modes.
Consider next the directions Xia. Using the fact that Xi
a∇iGb = i/2M2ab, as well as
the condition (2.2) to simplify the terms involving ∇iGj , one finds:
m2ijX
iaX
ja = eG
[
M2aa − i qaij
(
XiaG
j −X jaG
i)
+ q kai qajkG
iGj −Rijpq XiaX
jaG
pGq]
+[1
4M4
aa −(
qbijXiaX
ja +
1
2QabcM
2ca
)
Gb − 1
4
(
M2aa − 2GaGa
)
GbGb
−M2abGaG
b +1
4QabdQ
dac G
bGc +1
2QabcGaG
bGc]
. (5.3)
The conditions m2ijX
iaX
ja ≥ 0 represent in principle new additional constraints, which
are absent in theories involving only chiral multiplets. We believe however that these
conditions do not capture any extremely relevant new information, as they involve essen-
tially the complex partners of the Goldstone scalars, which are not expected to be very
dangerous 11.
We are now ready to use the conditions of flatness and stability at the stationary
point to find some algebraic constraints applying to any supergravity theory, and which
are simple enough to represent an interesting restriction on the functions G, hab and Xa
that specify the theory. Recall that the flatness condition is given by eq. (3.1) and takes
the form:
−3 +GiGi +1
2e−GGaGa = 0 . (5.4)
The necessary conditions for stability are more involved and correspond to requiring that
the expressions (5.2) and (5.3) should be semi-positive. This leads to two kinds of algebraic
conditions. Note however that (5.2) depends only on M2ab and not on qaij, and has a
simple tensor structure, whereas (5.3) depends both on M2ab and qaij, and has a much
more complicated tensor structure. As we already mentioned, we believe that only the
former leads to a truly significant condition so we will therefore take as our necessary
11This fact can be explicitly seen in the simplest case of models involving just one chiral and one
vector field. In this case there are only two mass eigenvalues, which must therefore be in one-to-one
correspondence with the two complex directions X and G′. One of these eigenvalues does not vanish and
is clearly associated to G′, whereas the other does vanish and is associated to X. It is then straightforward
to verify that the non-zero eigenvalue is exactly captured by (5.2), and that (5.3) is just proportional to
it.
16
condition for meta-stability the condition m2ijG
iGj ≥ 0, which implies:
RijpqGiGjGpGq ≤ 6 + e−G
[
−2GaGa + hcdhacihbdjGiGjGaGb
]
+ e−2G[
M2abG
aGb − 3
4QabcG
aGbGc (5.5)
−1
2
(
GaGa
)2+
1
4h k
ab hcdk GaGbGcGd
]
.
The flatness condition (5.4) and the stability constraint (5.5) represent a strong re-
striction on the theory. They generalize the results derived in [17] and [18] to the more
general case where both chiral and vector multiplets participate significantly to supersym-
metry breaking. In the following, to work out more explicitly the implications of these
constraints, we will mostly focus on the more tractable case where the gauge kinetic func-
tions are constant and diagonal, so that hab = g−2a δab, hai = 0 and haij = 0. In this
situation, one has also necessarily Qabc = 0. The flatness and stability conditions take in
this case the following simple form:
gijFiF j = 3m2
3/2 −1
2hab D
aDb ,
Rijpq FiF jF pF q ≤ 6m4
3/2 +(
M2ab − 2m2
3/2hab
)
DaDb − 1
2habhcdD
aDbDcDd .(5.6)
Besides this, it is important to remember that the values of the Fi and the Da are not
completely independent, but rather related by eq. (4.4). In the simpler situation considered
here, this relation implies, after using the flatness condition, that 12:
qaijFiF j =
1
2
[
M2ab + 4m2
3/2 hab
]
Db − 1
2habhcdD
bDcDd . (5.7)
Recall finally that the auxiliary fields Fi and Da are related to the complex directions
Gi and Xia by the relations Fi = −m3/2Gi and Da = iXi
aGi. This means that the Fi are
associated to the direction of supersymmetry breaking, whereas the Da are associated to
the scalar products between this direction and the gauge symmetry breaking direction.
Whenever the Da are small, the gauge symmetry breaking effects related to the vector
multiplets get weaker. More precisely, in the limit gaDa → 0 the eqs. (5.6) reduce to those
emerging in the absence of vector multiplets. This is not at all obvious, since in this limit
the D-terms still give a non-vanishing contribution to m2ij. As can be seen from (3.3)
this contribution is however given simply by g2aXaiXaj, and therefore vanishes along the
direction Gi, as gaGiXai = igaDa and this vanishes in the limit gaDa → 0. On the other
12Note that using the relations (5.7) it is possible to rewrite the quadratic term (M2ab−2m2
3/2hab)DaDb
in (5.6) as 2 qaijDaF iF j − 2 gijhab F iF jDaDb. One might then be tempted to further rewrite this as
2(Zijab − gijhab)F iF jDaDb, in terms of the new quantity Zijab = D−1(a ∇i∇jDb), which does however not
have any clear interpretation. This would allow us to rewrite the conditions (5.6) as Gξθηξηθ = m23/2 and
Rξθχψηξηθηχηψ ≤ 2/3 m43/2, in terms of the Goldstino vector ηξ = (iF i/
√3, Da/
√6) and some generalized
”metric” and ”curvature” tensors. Unfortunately, it does not seem to be possible to write the conditions
(5.7) in any illuminating way in terms of ηξ. One could then ignore (5.7) and treat all the components of
ηξ as independent variables, to define a weaker set of equations, in the same spirit as in [18], although we
will not follow this direction here.
17
hand, this limit implies also some constraints among the chiral multiplets. If the masses
of the vectors are small (gaMab → 0), then the kinematical relation between Da and Fi
implies gaXiaFi → 0, whereas if they are large (gaMab → ∞), the dynamical relation (5.7)
between Da and Fi implies instead gaqaijFiF j → 0.
Finally we would like to briefly comment on the special case of models involving just
one chiral multiplet and one vector multiplet. In this case the physical mass matrix has
just 1 entry and the condition for it to be positive is given exactly by (5.5). In this case,
the symbol ≤ in (5.5) and (5.6) can therefore be replaced by <, and the condition is not
only necessary but also sufficient.
6 Analysis of the constraints
The flatness and stability conditions (5.6) represent our main result regarding the con-
straints put on gauge invariant supergravity theories by the requirement that they should
admit a phenomenologically and cosmologically viable vacuum. Along this section we will
address the problem of working out more concretely the implications of these constraints.
We assume for simplicity that the gauge group consists of only Abelian U(1) factors. In
such a case, it is always possible to choose the basis of vector fields so that the mass matrix
of the gauge bosons is diagonal, M2ab = M2
aδab. The gauge kinetic function is assumed to
be diagonal as well, and therefore the gauge couplings can be written as hab = g−2a δab. We
can then choose the special parametrization of scalar fields that corresponds to normal
coordinates for the Kahler geometry, in which the Kahler metric becomes δIJ . Similarly,
we can rescale the vector fields to define flat indices also for vector indices, in such a way
that the vector metric becomes just δAB . This rescaling simply amounts to including a
factor ga for each vector index a, and one has for instance DA = gaDa and M2A = g2
aM2a .
In this way, no explicit dependence on ga is left in the formulae.
It is convenient to introduce new variables, obtained by suitably rescaling the auxiliary
fields with the gravitino mass as follows:
f I =F I
√3m3/2
, dA =DA
√6m3/2
. (6.1)
Note that the flat capital indices of these new fields are raised and lowered with the
diagonal metrics δIJ and δAB . In addition, we also find convenient to introduce the
following new quantities, defined by rescaling the Killing vectors and its derivatives by
the vector masses:
vIA =
√2XI
A
MA, TAIJ =
qAIJ
MA. (6.2)
Finally, we shall introduce the following dimensionless parameters, measuring the hierar-
chies between the scales of gauge symmetry and supersymmetry breaking:
ρA =MA
2m3/2. (6.3)
18
In terms of the quantities we have just introduced, the conditions (5.6) can be rewritten
in the following form:
δIJ fIf J = 1 −
∑
Ad2A ,
RIJP Q fIf JfPf Q ≤ 2
3+
4
3
∑
A
(
2 ρ2A − 1
)
d2A − 2
∑
A,Bd2Ad
2B ,
(6.4)
whereas the dynamical relation (4.4), kinematical relation (4.1) and kinematical bound
(4.2) read now:
dA =
√
3
2
ρA TAIJ fIf J
ρ2A − 1/2 + 3/2 δIJ f
If J, (6.5)
dA = iρAvIAfI , (6.6)
|dA| ≤ ρA
√
δIJ fIf J . (6.7)
Notice that the bound (6.7) and the flatness condition (6.4) imply, together, the following
limits on the sizes of the auxiliary fields:
√
δIJ fIf J ∈
[
1√
1 +∑
Bρ2B
, 1
]
,√
∑
Ad2A ∈
[
0,
√
∑
Bρ2B
1 +∑
Bρ2B
]
. (6.8)
The size of the dA auxiliary field is therefore limited by the kinematical bound for small
ρA and by the flatness condition for large ρA.
To derive the implications of the constraints (6.4), one should take into account the fact
that f I and dA are not independent variables. There are then several possible strategies
that one can follow. A first possibility is to use the the dynamical relation (6.5) to write
the dA in terms of the f I . One can then consider the parameters RIJP Q, MA and TAIJ as
fixed and take as free variables the f I . In this way the equations have a reasonably simple
tensor structure, but they become of higher order in the variables. As a consequence, they
can in principle be solved only in special limits on the values of the parameters, like for
instance the limits of heavy and light vectors, where the relation (6.5) simplifies and the
effects induced by the dA represent only small corrections. A second possibility is to use
the kinematical relation (6.6) to write the dA in term of the f I . One can then consider
the parameters RIJP Q and XIA as fixed, and the quantities f I as free variables. The
equations have then the same structure as for a theory with only chiral multiplets, with at
most quartic terms, but their tensor structure is substantially more involved. Therefore
in general the equations cannot be solved, and this type of analysis is efficient only in
special cases, where RIJP Q and/or XIA have special properties. A third possibility is to
impose only the kinematical bound (6.7) to restrict the values of the dA in terms of the
f I . One can then consider the parameters RIJP Q and MA as fixed, and the quantities
f I and dA as free variables subject only to a bound. By doing so, one clearly looses a
substantial amount of information, and the resulting conditions will therefore be weaker
than the original ones, although still necessary. This is similar in spirit to what was done
in [18] to study cases with only chiral multiplets but with a Kahler geometry leading to a
complicated structure of the Riemann tensor RIJP Q.
19
It is clear that switching from the dynamical relation (6.5) to the kinematical relation
(6.6) and finally to the kinematical bound (6.7) represents a gradual simplification of the
formulae, which is also accompanied by a loss of information. As a consequence, the three
different types of strategies we just described, which are based on the use of these three
different relations, will be tractable over an increasingly larger domain of parameters,
but this will be accompanied by a gradual weakening of the implied constraints. In the
following sections we will perform more explicitly these different analyses and see what
kind of information can be obtained with each of them.
7 Exploiting the dynamical relation between dA and fI
As already discussed in section 4, whenever the masses MA of the vector multiplets are all
much larger or much smaller than m3/2, that is ρA ≫ 1 or ρA ≪ 1, the situation simplifies.
In those limits (and assuming that the quantities TAIJ remain finite) the auxiliary fields
dA become small, and the dynamical relation (6.5) simplifies. It becomes then convenient
to use it and to apply the first strategy outlined in section 6.
For the purpose of this section, it is useful to rewrite the conditions (6.4) in a slightly
different form, which exhibits in a better way the effect of the vector multiplets with
respect to the situation involving only chiral multiplets. To do so, we define new rescaled
variables in such a way to reabsorb the non-trivial right-hand side of the flatness condition,
and to transfer all the non-trivial effect of the vector multiplets into the right-hand side
of the stability condition. This can be done by introducing the following new quantities:
zI =f I
√
1 −∑
Ad2A
. (7.1)
Using these new variables zI instead of the f I , the flatness and stability conditions (6.4)
can be rewritten as:
δIJ zIzJ = 1 ,
RIJP Q zIzJzP zQ ≤ 2
3K(d2
A, ρ2A) ,
(7.2)
where
K(d2A, ρ
2A) = 1 + 4
∑
Aρ2Ad
2A −
(∑
Ad2A
)2
(
1 −∑
Bd2B
)2 . (7.3)
Due to the non-linearity of the change of variable (7.1), the dynamical relation (6.5)
becomes more complicated and does not allow us to explicitly express the variables dA in
terms of the new zI . Similarly, the kinematical relation (6.6) and the kinematical bound
(6.7) also get modified. More explicitly, these three relations between auxiliary fields take
20
now the following forms:
dA
1 + ρ2A − 3/2
∑
Bd2B
1 −∑
Bd2B
=
√
3
2ρA TAIJ z
IzJ , (7.4)
dA√
1 −∑
Bd2B
= iρAvIAzI , (7.5)
|dA|√
1 −∑
Bd2B
≤ ρA . (7.6)
This formulation of the constraints becomes identical to the previous one for small dA. It
presents however some advantages. Note in particular that in the limit dA ≪ 1 the new
variables zI defined by (7.1) approximately coincide with the f I , and can be therefore
efficiently used to study the leading order effects due to vector multiplets.
We will now show that for the cases we are going to study in this section, the conditions
(7.2) can be rewritten in the form:
δIJ zIzJ = 1 ,
RIJP Q zIzJzP zQ ≤ 2
3.
(7.7)
These equations have the same functional form as the ones found in the case with only
chiral multiplets. However the relevant variables zI are rescaled with respect to the original
variables fI , and the coefficients RIJP Q are shifted with respect to the components of the
Riemann tensor RIJP Q.
7.1 Heavy vector limit
In the limit ρA ≫ 1, the relation (7.4) implies that the dA are small and given by:
dA ≃√
3
2ρ−1
A TAIJ zIzJ . (7.8)
In the small dA limit the function K in (7.3) can then be simplified by keeping only the
leading term and evaluating it by using eq. (7.8). This results in a quartic dependence on
the variables zi. One can then rewrite the conditions in the form (7.7), with an effective
curvature tensor given by:
RIJP Q ≃ RIJP Q − 2∑
A
(
TAIJ TAPQ + TAIQ TAP J
)
. (7.9)
In addition, note that since dA ≃ 0, the kinematical relation (7.5) implies them constraints
vIAzI ≃ 0.
These formulae show that in the limit in which the vector multiplets are heavy, they
give two kinds of effects. On one hand, they induce a correction to the Kahler curvature
for the chiral multiplets. On the other hand, they reduce the number of relevant variables
zI coming from the chiral multiplets, as some of the directions are associated to the would-
be Goldstone chiral multiplets that are absorbed by the vector multiplets. It is clear from
21
the form of the new tensor (7.9) that the correction on the curvature induced by the
vector multiplets is negative, so the net effect is (compared with the situation with just
chiral fields) to help in fulfilling the bound (7.7). Note that the correction in (7.9) is not
necessarily small and can be of O(1). Actually, even for TAIJ ∼ O(1), one has zI ∼ O(1)
and dA ∼ O(ρ−1A ), so that the ratio dA/zI ∼ O(ρ−1
A ) is still small, as assumed.
We can understand this result also from a slightly different point of view. We see
from (7.9) that the correction induced by a heavy vector field to the Kahler curvature
is negative and of order M−2A . This is by assumption small compared to m−2
3/2, but it
can still be large compared to 1 (in Planck units). The original curvature R of the
chiral multiplet theory is instead of order Λ−2, where Λ is the scale of the most relevant
higher-dimensional operator. One must certainly have Λ ≫ m3/2, and gravitational effects
correspond to Λ ∼ 1. The effects due to vector multiplets can thus really compete with
the curvature effects due to chiral multiplets, or even dominate over them. In particular,
when the original curvature effects are of gravitational origin, the vector multiplet effects
become significant whenMA ∼ 1. This is typically what happens for spontaneously broken
anomalous U(1) symmetries in string models.
As we explained in section 4, in the limit ρA ≫ 1 it should be possible to integrate
out the vector multiplets and reinterpret their net effect through a correction to the
Kahler potential of the chiral multiplets. It is interesting to verify that such a procedure
indeed reproduces (7.9). To do so, we can use superfields and follow the same steps as in
subsection 4.1. Recall in particular that the correction to the potential G was shown to
be equal to ∆G ≃ −∑
AM−2A G2
A. Since in flat indices the Riemann tensor is given by the
fourth derivative of the potential G, the corresponding correction to the curvature is:
∆RIJP Q ≃ −∑
AM−2A ∇I∇J∇P∇Q(G2
A) . (7.10)
Taking into account that ∇I∇JGA = 0 and ∇I∇JGA = −qAIJ , one recovers then the
second term in the expression (7.9). Notice however that this simplified derivation of
(7.9) is valid only under the assumption that the correction is small, whereas as already
emphasized, (7.9) is actually valid also in more general situations where the correction
can be sizable.
7.2 Light vector limit
In the limit ρA ≪ 1, the bound (7.6) and the relation (7.4) imply that the dA are small
and given by
dA ≃√
3
2ρA TAIJ z
IzJ . (7.11)
Again, as in the previous subsection, we can simplify the function K by keeping only the
quartic term in zi. This allows us to finally rewrite the conditions (7.2) in the form (7.7),
with an effective curvature tensor now given by:
RIJP Q ≃ RIJP Q − 2∑
Aρ4A
(
TAIJ TAPQ + TAIQ TAP J
)
. (7.12)
22
Therefore, the net effect of the vector multiplets is as before to reduce the Kahler curvature
(with respect to the case with only chiral multiplets). In addition, the number of relevant
directions in field space is again reduced, since the kinematical bound (7.6) together with
the dynamical relation (7.11) imply the m constraints TAIJ zIzJ ≃ 0.
7.3 Arbitrary vector masses
One may wonder whether it is possible to generalize the analyses done in subsections 7.1
and 7.2 to the case of arbitrary vector masses. In particular, the crucial question is whether
the effect of vector multiplets could again be essentially encoded into a modification of
the Riemann tensor. It is clear from the structure of the equations (7.2) that in general
this is not going to be the case. However, we will see that the original constraints imply
some other weaker constraints, which have indeed this form but contain less information
except in the cases of large or small masses.
In the case of only one vector multiplet, the flatness condition guarantees that the
numerator in (7.4) never vanishes and stays always positive. Moreover, the kinematical
bound (7.6) implies that it reaches its minimal value for d = ρ/√
1 + ρ2. Using this result,
we can then get an upper bound for the quantity |d|/(1−d2) by evaluating the numerator
at its minimum. In this way we find that:
|d|1 − d2
≤√
3
2
ρ(
1 + ρ2)
1 + ρ2/2 + ρ4
∣
∣TIJzIzJ∣
∣ . (7.13)
We can now find an upper bound to the function K by neglecting the negative term in
(7.3) and using the bound (7.13). We can finally substitute this upper bound for K in the
conditions (7.2) and deduce a simpler but weaker set of the conditions. These have once
again the form (7.7), where now:
RIJP Q = RIJP Q − 2 ǫ(ρ)(
TIJ TPQ + TIQ TP J
)
, (7.14)
with a function ǫ(ρ) given by:
ǫ(ρ) =ρ4(
1 + ρ2)2
(
1 + ρ2/2 + ρ4)2 . (7.15)
This expression is valid for arbitrary ρ and in the limits ρ ≫ 1 and ρ ≪ 1 it correctly
reproduces (7.9) and (7.12).
In the more general case of several vector multiplets, the situation is more complicated,
due to the fact that the numerator in the left-hand side of the dynamical relation (7.4) can
vanish 13. For this reason it is no longer possible to find a simple bound of the type (7.13)
valid for all values of the parameters ρA. One can however still derive a useful bound in
a large domain of the space of values of the parameters ρA. To see which domain should
be considered, note that the kinematical bound implies that:
1 + ρ2A − 3
2
∑
Bd2B ≥
1 + ρ2A −
∑
Bρ2B/2 + ρ2
A
∑
Bρ2B
1 +∑
Bρ2B
. (7.16)
13It is however clear that whenever this numerator becomes small, also the right-hand side of the equation
must do so.
23
The right-hand side of (7.16) is positive if
ρ2A ≥ 1
4
(
−1 − 2∑
B 6=Aρ2B +
√
4(∑
B 6=Aρ2B
)2+ 12
∑
B 6=Aρ2B − 15
)
. (7.17)
The function in the right hand side of the inequality (7.17) vanishes when∑
B 6=Aρ2B = 2.
For lower values,∑
B 6=Aρ2B < 2, the function is negative and the condition is thus always
satisfied. On the other hand, for larger values,∑
B 6=Aρ2B > 2, the function is positive and
there is thus a non-trivial restriction on the parameters. It actually turns out that in this
case this function grows monotonically up to the maximal value of 1/2, so if ρ2A > 1/2 the
condition (7.17) is automatically satisfied. This shows that the numerator in (7.16) can
vanish only in very special valleys of parameters space, where some of the ρA are small
and some other are large. On the contrary, if all of them are either not too small or no
too large, then this problem does not appear. We can then restrict our analysis to the
following two regions of parameter space:
I> ={
~ρ∣
∣
∣ρ2
A ≥ 1
2,∀A
}
, I< ={
~ρ∣
∣
∣ρ2
A ≤ 2 ,∀A}
. (7.18)
In the domain I< ∪ I> one can then deduce the following simple upper bound for (7.4):
|dA|1 −∑
Bd2B
≤√
3
2
ρA
(
1 +∑
Bρ2B
)
1 + ρ2A −
∑
Bρ2B/2 + ρ2
A
∑
Bρ2B
∣
∣TAIJzIzJ∣
∣ . (7.19)
This is the obvious generalization of the bound (7.13) for the case of more than one vector
multiplet. One can then proceed as before, and use this to derive an upper bound for the
function K, which can then be substituted in the original conditions to deduce a simpler
but weaker set of constraints. In this way one finds once again (7.7), with an effective
curvature of the form
RIJP Q = RIJP Q − 2∑
Aǫ(ρA)(
TAIJ TAPQ + TAIQ TAP J
)
, (7.20)
where ǫ(ρA) is now given by
ǫ(ρA) =ρ2
A
(
1 +∑
Bρ2B
)2
(
1 + ρ2A −
∑
Bρ2B/2 + ρ2
A
∑
Bρ2B
)2 . (7.21)
This expression correctly reproduces (7.9) and (7.12) in the limits ρA ≫ 1 and ρA ≪ 1. It
also reproduces (7.15) in the case of a single vector multiplet, since in that case I< ∪ I>coincides with the whole parameter space.
7.4 Simple scalar geometries
It was shown in [17, 18] that in the case of theories with only chiral multiplets the condi-
tions of flatness and stability could be solved exactly for certain particular classes of scalar
manifolds leading to a simple structure for the Riemann tensor. This is for instance the
case when the scalar manifold factorizes in a product of one-dimensional scalar manifolds
(factorizable scalar manifolds), or when it is a symmetric coset group manifold of the
24
form G/H (symmetric scalar manifolds). It is then illustrative and physically interesting
to apply the results derived in this subsection to these simple examples.
For factorizable manifolds, the metric is diagonal and therefore the structure of the
Riemann tensor simplifies to RIJP Q = RI δIJP Q. Assuming then also that the Killing
potentials are separable, so that TAIJ = qAI/MA δIJ , the corrected effective curvature
appearing in (7.20) takes the following simple form:
RIJP Q = RI δIJP Q − 2∑
Aǫ(ρA)qAIqAP
M2A
(
δIJ δPQ + δIQ δP J
)
. (7.22)
This is no longer separable, but the problem can however still be solved exactly. Indeed,
the flatness and stability conditions (7.7) reduce to
∑
I |zI |2 ≃ 1 ,
∑
I,JRIJ |zI |2 |zJ |2 ≤ 2
3,
(7.23)
where the n-dimensional matrix RIJ is given by
RIJ = RI δIJ − 4∑
Aǫ(ρA)qAI qAJ
M2A
. (7.24)
The values of the variables zI minimizing the stability condition subject to the flatness
constraint are easily found to be |zI |2 =∑
J(R−1)IJ/∑
R,S(R−1)RS . Substituting these
values into the stability bound, one finds then the following necessary condition:
∑
I,J(R−1)IJ ≥ 3
2. (7.25)
Whenever the correction to the curvature induced by the vector multiplets is small, the
inverse of the matrix (7.24) is easily found and the above curvature constraint simplifies
to:
∑
Aǫ(ρA)(qAIR
−1I )2
M2A
>∼1
4
(3
2−∑
IR−1I
)
. (7.26)
It is then clear that in a case where∑
I R−1I is only slightly lower than 3/2, the effect of
vector multiplets can help in satisfying the bound.
For maximally symmetric scalar manifolds, the Riemann tensor is related to the
metric and takes the form RIJP Q = Rall/2 (δIJ δPQ + δIQδP J), where Rall is an over-
all curvature scale. Assuming then as before that the Killing potentials are diagonal,
TAIJ = qAI/MA δIJ , the corrected effective curvature takes the form:
RIJP Q =(Rall
2− 2∑
Aǫ(ρA)qAI qAP
M2A
)(
δIJ δPQ + δIQ δP J
)
. (7.27)
This is still of the maximally symmetric form, and the flatness and stability conditions
(7.7) can then be written in the form:
∑
I |zI |2 ≃ 1 ,
∑
I,J∆RIJ |zI |2 |zJ |2 ≥ Rall −2
3,
(7.28)
25
where the n-dimensional matrix ∆RIJ is given by
∆RIJ = 4∑
Aǫ(ρA)qAI qAJ
M2A
. (7.29)
The values of the zI that optimize the stability bound taking into account the flatness con-
dition are again easily found, and read in this case |zI |2 =∑
J(∆R−1)IJ/∑
R,S(∆R−1)RS .
Substituting these values into the stability bound, one finds then the following necessary
condition:
∑
R,S(∆R−1)RS ≤(
Rall −2
3
)−1. (7.30)
This means that whenever Rall ≥ 2/3, there need to be vector fields with sufficiently low
mass and large charges to avoid unstable modes. For instance, in the particular case with
a single vector multiplet, the necessary condition is just qmax/M ≥ 1/2√
Rall − 2/3.
8 Exploiting the kinematical relation between da and fI
Whenever the Riemann tensor and the charge matrix have simple tensor structures, the
kinematical relation (6.6) can become useful in the task of solving the constraints (6.4).
This is the case, for instance, when the scalar manifold is factorizable and the isometries
that are gauged are each aligned along a single one-dimensional submanifold. In such a
situation the Riemann tensor is diagonal and has the form RIJP Q = RI δIJP Q and each
Killing vector has a single non-vanishing component, XIA = XI δI
A. In this situation it is
convenient to use the second strategy outlined in section 6. In this case the number of
chiral and vector multiplets are equal: m = n and we can then identify the chiral and
vector indices I, J, . . . and A,B, . . . , keeping in mind that simpler cases with less vector
multiplets can be described by simply setting some of the XI to 0 14.
The crucial simplification in this case is that, due to the kinematical relation (4.1),
each vector auxiliary field dI is proportional to the corresponding chiral auxiliary field fI .
For each one-dimensional subspace of the scalar manifold, the ratio of the two auxiliary
fields is actually fixed by the mass of the corresponding vector field: |dI |/|fI | = ρI . The
orientation of the Goldstino direction along the n one-dimensional subspaces of the scalar
manifold is instead arbitrary. It can be parametrized by n variables xi related to the
absolute sizes of |fi| and |di| in each of these subspaces. A convenient choice is to define:
xI =√
|fI |2 + |dI |2 . (8.1)
14It should be emphasized that this simple example where the space factorizes in a number of sectors
with one pair of chiral and vector multiplets each is rather restrictive. Indeed, the requirement of gauge
invariance of G severely restricts the possible form of the superpotential W . In the case where both K
and W are gauge invariant, W can only be a constant, implying that non-zero fI and dI can be induced
only through a non-trivial K. A more general possibility is that K and W have gauge variations that
compensate each other. This fixes however almost uniquely the form of W in terms of the form of K, and
one finds again a very similar situation.
26
The flatness and stability conditions can then be rewritten in the following very simple
form:
∑
Ix2I = 1 ,
∑
I,JRIJ x2I x
2J +
2
3
∑
IEI x2I ≤ 2
3,
(8.2)
where
RIJ =RI δIJ + 2 ρ2
Iρ2J
(
1 + ρ2I
)(
1 + ρ2J
) , EI =2 ρ2
I
(
1 − 2 ρ2I
)
(
1 + ρ2I
) . (8.3)
The problem can now be solved along the same lines as in [17, 18]. The constraint (8.2) can
be interpreted as an upper bound on the function f(xI) =∑
I,J RIJ x2I x
2J +2/3
∑
I EI x2I ,
where the real variables xI range from 0 to 1 and are subject to the constraint∑
K x2K = 1.
In particular, the inequality (8.2) implies that fmin < 2/3, where fmin is the minimum value
of f(xI) within the allowed range for the xI . Finding fmin is a constrained minimization
problem which can be solved in the standard way using Lagrangian multipliers. It is
straightforward to show that the values of the variables at the minimum are given by
x2I =
1∑
P,QR−1PQ
[
∑
J R−1IJ − 1
3
∑
J,K,LR−1IJ R
−1KL
(
EJ − EL
)
]
. (8.4)
The condition fmin < 2/3 then implies that the constraint on the curvatures takes the
form:∑
I,JR−1IJ
(
1 − EI
)
− 1
6
∑
I,J,P,QR−1IJ R
−1PQ
(
EJ − EP
)
EQ ≥ 3
2. (8.5)
The expression (8.5) represents the main result of this section. It correctly reduces to
the condition∑
I R−1I > 3/2 in the limit where all the gauge couplings are switched off.
The matrices RIJ are positive definite and reduce to RIδIJ at zero coupling, whereas the
parameters EI can be either positive or negative but vanish at zero coupling.
Using the expression of RIJ , the inequality (8.5) can actually be rewritten more ex-
plicitly in terms of the inverse curvatures R−1I . The result takes the form:
∑
IαI R−1I +
∑
I,JβIJ R−1I R−1
J +∑
I,J,KγIJK R−1I R−1
J R−1K ≥ 3
2, (8.6)
where the coefficients αI , βIJ and γIJK are positive and depend on the mass parameters
ρI as follows:
αI = 1 + 4 ρ6I ,
βIJ =8
3
(
ρ2I − ρ2
J
)2(1 − ρ2
I − ρ2J + ρ4
I + ρ4J − 2ρ2
Iρ2J + 2 ρ4
Iρ2J + 2 ρ4
Jρ2I + ρ4
Iρ4J
)
,
γIJK =16
3
(
ρ2I − ρ2
J
)2(ρ2
I − ρ2K
)2(ρ2
J − ρ2K
)2. (8.7)
Note that in the limit ρI → 0 the constraint (8.6) reduces to the correct condition∑
I R−1I ≥ 3/2 that was found in [17, 18] for the case of theories with only chiral mul-
tiplets. Actually, as all the coefficients αI , βIJ and γIJK are positive and, in particular,
αI > 1 then the condition (8.6) is less stringent than the condition found in that case.
Note also that for large ρI the bound is trivially satisfied.
27
8.1 One pair of chiral and vector multiplets
In the simplest case of models involving 1 chiral multiplet and 1 vector multiplet, the
situation is particularly simple. The flatness condition fixes x = 1. The auxiliary fields
can then be parametrized as |f | = cos δ and |d| = sin δ, where the angle δ is completely
fixed by the parameter ρ as ρ = tan δ. The condition (8.6) involves in this case only the
first term so is linear in the inverse curvatures, and it implies the constraint:
R ≤ 2
3
(
1 + 4 tan6δ)
. (8.8)
This shows that the stability condition can always be satisfied for sufficiently large values
of δ; more precisely, one needs δ ∈ [δmin, π/2], where
δmin =
0 , R <2
3,
arctan[3
8
(
R− 2
3
)]1/6, R >
2
3.
(8.9)
As expected, the bound R < 2/3 found in the limit of vanishing δ (in which the effect of
the vector multiplet is negligible, and we recover the result found for just one chiral field),
get corrected for non-vanishing δ. Actually when the effect of the vector multiplet starts
becoming important the bound gets milder and eventually trivializes for large δ. Note
however that the first correction appears only at sixth order in δ.
8.2 Two pairs of chiral and vector multiplets
The next-to-simplest case is the case of models involving 2 chiral multiplets and 2 vector
multiplets. The solution of the flatness condition can in this case be parametrized in terms
of an arbitrary angle θ so that x1 = cos θ and x2 = sin θ. The auxiliary fields can then
be written as f1 = cos θ cos δ1, d1 = cos θ sin δ1, f2 = sin θ cos δ2, d2 = sin θ sin δ2, where
the angles δ1,2 are completely fixed by ρ1,2 = tan δ1,2. The condition (8.6) involves in this
case the first two terms, which are linear and quadratic in the curvatures, and implies a
rather complicated constraint involving the curvatures R1,2 and the angles δ1,2. From its
structure, and the fact that the relevant coefficients αI and βIJ are positive, it is however
clear that this constraint is milder than the constraints that would arise for each pair of
chiral and vector multiplets on its own. For example, in the particular case where the two
sectors are identical one would find the same constraint as for a single basic sector with a
pair of chiral and vector multiplet, but with an effective curvature reduced by a factor of
2, as is also the case when only chiral multiplets are present.
8.3 Several pairs of chiral and vector multiplets
In the more general case of models involving an arbitrary number of pairs of chiral and
vector multiplets, the situation is even more complicated. Nevertheless it is interesting
to point out that for situations where the parameters ρI satisfy certain properties, it is
possible to derive a condition that is linear, rather than cubic, in the inverse curvatures
28
R−1I . More precisely, this is the case in the limit where all the ρI are such that ρ2
I ≤ 1/2.
In that case, the quantity EI is positive definite, so then it is possible to get a new
(but weaker) condition by disregarding the term involving EI from the conditions (8.2).
Following the same procedure that was used to derive eq. (8.6), we get the necessary
condition:
∑
I
(
1 + ρ2I
)2R−1
I ≥ 3
2. (8.10)
This means that in this domain of parameters, namely ρ2I ≤ 1/2, and the approximation
considered here, the net effect of the vector multiplets is to effectively reduce the curvatures
RI by a factor (1 + ρ2I)
−2, that is, RI =(
1 + ρ2I
)−2RI .
9 Exploiting the kinematical bound between da and fI
In more general cases, for which the analyses of sections 7 and 8 cannot be applied, one
may try to work out the implications of the constraints (6.4) by applying the third strategy
outlined in section 6. This consists in considering both the f I and the dA as independent
variables, constrained only by the kinematical bound (6.7). This approach is simple and
can be worked out in full generality; however it clearly ignores a substantial amount of
information concerning the actual dynamical and kinematical relations between the f I
and the dA. The resulting implications will therefore be weaker than the ones derived
in the previous sections. Note in this respect that the kinematical bound represents a
significant constraint only when ρA <∼ 1, whereas it becomes trivial when ρA ≫ 1. We
therefore expect that the condition resulting from this analysis will be stronger for small
ρA and will become weaker and weaker for increasing ρA.
For this analysis, it is convenient to use the new variables zI defined in (7.1) instead
of the fI , and also similarly rescaled variables ǫA instead of the dA. More precisely, we
consider the following change of variables:
zI =f I
√
1 −∑
Ad2A
, ǫA =dA
√
1 −∑
Bd2B
. (9.1)
Using these new variables, the flatness and stability conditions (7.2) can be rewritten in
the simple form:
δIJ zIzJ = 1 ,
RIJP Q zIzJzP zQ ≤ 2
3K(ǫ2A, ρ
2A) ,
(9.2)
where now
K(ǫ2A, ρ2A) = 1 + 4
∑
Aρ2A ǫ
2A − 4
∑
A,Bǫ2A ǫ
2B + 4
∑
A,Bρ2A ǫ
2A ǫ
2B , (9.3)
and the kinematical bound (6.7) becomes simply:
|ǫA| ≤ ρA . (9.4)
29
It is now clear from (9.2) that the most favorable situation is when the variables
zI minimize the function RIJP Q zIzJzP zQ and the variables ǫA maximize the function
K(ǫ2A, ρ2A). The maximization of K(ǫ2A, ρ
2A) with respect to ǫA should be done taking into
account the bound (9.4):
Kbest(ρ2A) = max
{
K(ǫ2A, ρ2A)∣
∣
∣ǫ2A ≤ ρ2
A
}
. (9.5)
The computation of this quantity is complicated by the fact that the function K(ǫ2A, ρ2A)
depends on all the variables ǫ2A, and that its maximum may lie inside region ǫ2A < ρ2A
or at the boundary ǫ2A = ρ2A, depending on the values of the parameters ρ2
A. It is then
a bit laborious, although straightforward, to characterize the constrained maximum of
the function K(ǫ2A, ρ2A) over the full range of the parameters ρ2
A. Once this maximal
value Kbest(ρ2A) has been found, one can substitute it in the stability condition to get the
following necessary conditions:
δIJ zIzJ = 1 ,
RIJP Q zIzJzP zQ ≤ 2
3,
(9.6)
where
RIJP Q = Kbest(ρ2A)−1RIJP Q . (9.7)
Again, from the structure of this constrain we see that the net effect of the vector multiplets
is to reduce the effective curvature that is perceived by the chiral multiplets, in this case
by the multiplicative factor K−1best, (which is clearly smaller than 1). Unfortunately, it does
not seem to be possible to find a closed expression for Kbest in general. We will thus first
examine in detail the simplest models with 1 and 2 vector multiplets, and then discuss
what can be said about the general case with n vector multiplets.
9.1 One vector field
In the presence of one vector field, the function (9.3) takes the form
K(ǫ2, ρ2) = 1 + 4 ρ2 ǫ2 + 4(
ρ2 − 1)
ǫ4 . (9.8)
The maximum of this function of ǫ2 within the region [0, ρ2] sits either at the stationary
point ρ2/(2 − 2ρ2) or at the boundary point ρ2, depending on the value of ρ2. More
precisely, if we define the two domains
I1 =[
0,1
2
]
, I2 =[1
2,+∞
[
, (9.9)
one finds:
ǫ2best =
1
2
ρ2
1 − ρ2, ρ2 ∈ I1 ,
ρ2 , ρ2 ∈ I2 .
(9.10)
30
The corresponding maximal value of the function K is then:
Kbest =
1 − ρ2 + ρ4
1 − ρ2, ρ2 ∈ I1 ,
1 + 4 ρ6 , ρ2 ∈ I2 .
(9.11)
Note that when a single chiral multiplet is present, the kinematical relation fixes ǫ2 = ρ2,
and one has to take the second branch of the expression (9.11) for any value of ρ. On
the other hand, when two or more chiral multiplets are present, one can in principle have
ǫ2 < ρ2, and one needs to consider the first branch of (9.11) for small values of ρ.
9.2 Two vector fields
In the presence of two vector fields, the function (9.3) takes the form
K(ǫ21,2) = 1 + 4(
ρ21 ǫ
21 + ρ2
2 ǫ22
)
+ 4(
ρ21 − 1
)
ǫ41 + 4(
ρ22 − 1
)
ǫ42
+ 4(
ρ21 + ρ2
2 − 2)
ǫ21 ǫ22 . (9.12)
The problem is symmetric, and we can thus assume without loss of generality that ρ21 ≤ ρ2
2.
The maximum of the function (9.12) with respect to (ǫ21,ǫ22) within the region [0, ρ2
1]×[0, ρ22]
sits at the points (ρ21/(2−2ρ2
1), 0) and (ρ21, ρ
22+ρ2
1(ρ21+ρ2
2−2)/(2−2ρ22)), or at the boundary
points (ρ21, 0) and (ρ2
1, ρ22). More precisely, the relevant domains turn out to be:
IA =
{
(ρ21, ρ
22)∣
∣
∣ρ21 ∈
[
0,1
2
]
, ρ22 ∈
[
0, ρ21
]
}
, (9.13)
IB =
{
(ρ21, ρ
22)∣
∣
∣ρ21 ∈
[1
2, 2]
, ρ22 ∈
[
0,2 − ρ2
1
1 + ρ21
ρ21
]
}
, (9.14)
IC =
{
(ρ21, ρ
22)∣
∣
∣ρ21 ∈
[1
2,
2√7
]
, ρ22 ∈
[2 − ρ21
1 + ρ21
ρ21, ρ
21
]
}
, (9.15)
ID =
{
(ρ21, ρ
22)∣
∣
∣ρ21 ∈
[ 2√7, 2]
, ρ22 ∈
[2 − ρ21
1 + ρ21
ρ21,
1 − ρ21 +
√
1 + 14ρ21 − 7ρ4
1
4
]
}
, (9.16)
IE =
{
(ρ21, ρ
22)∣
∣
∣ρ21 ∈
[ 2√7, 2]
, ρ22 ∈
[1 − ρ21 +
√
1 + 14ρ21 − 7ρ4
1
4,+∞
[
}
, (9.17)
IF =
{
(ρ21, ρ
22)∣
∣
∣ρ21 ∈
[
2,+∞[
, ρ22 ∈
[
0,+∞[
}
, (9.18)
and the constrained maximum is located at the following points:
(
ǫ21, ǫ22
)
best=
(1
2
ρ21
1 − ρ21
, 0)
,(
ρ21, ρ
22
)
∈ IA ,
(
ρ21, 0)
,(
ρ21, ρ
22
)
∈ IB ,
(
ρ21,
1
2
ρ22 + ρ2
1(ρ21 + ρ2
2 − 2)
1 − ρ22
)
,(
ρ21, ρ
22
)
∈ IC ∪ ID ,(
ρ21, ρ
22
)
,(
ρ21, ρ
22
)
∈ IE ∪ IF .
(9.19)
31
The corresponding maximal value of the function K(ǫ21,2) is:
Kbest =
1 − ρ21 + ρ4
1
1 − ρ21
,(
ρ21, ρ
22
)
∈ IA ,
1 + 4 ρ61 ,
(
ρ21, ρ
22
)
∈ IB ,
1 − ρ22 + ρ4
2 + ρ21
(
ρ21 + 2
)
ρ22
(
ρ22 − 2
)
− 2ρ61ρ
22 + ρ8
1
1 − ρ22
,(
ρ21, ρ
22
)
∈ IC ∪ ID ,
1 + 4(
ρ61 + ρ6
2
)
+ 4ρ21ρ
22
(
ρ21 + ρ2
2 − 2)
,(
ρ21, ρ
22
)
∈ IE ∪ IF .
(9.20)
9.3 Arbitrary number of vector fields
For a larger number of vector multiplets, the maximization problem becomes more and
more complex, and it does not seem to be possible to find the general solution. Never-
theless, it turns out to be possible to make some quantitative analysis also in this more
general case, by either making some assumptions on the parameters ρA or by weakening
the conditions.
In the particular case where all the parameters ρA are such that ρA > 1 it is clear from
the form of the expression (9.3) that the constrained maximum of the function K(ǫ2A, ρ2A)
is located at the corner of the parameter space where all the ǫ2a are maximal, and therefore
ǫ2A best = ρ2A. In terms of the variables dA, this means that the preferred situation is the
one in which d2A = ρ2
A/(1 +∑
B ρ2B). Note that this value is in general less than the
maximal value that each d2A is individually allowed to take by the kinematical bound, but
their sum is instead equal to the maximal value∑
A d2A =
∑
A ρ2A/(1+
∑
B ρ2B) it is allowed
to take. In other words, the optimal situation is realized in one particular direction in the
space of the dA such that∑
A d2A is maximal. The corresponding value of the function K
is in this case:
Kbest = 1 + 4∑
Aρ4A − 4
∑
A,Bρ2A ρ
2B + 4
∑
A,Bρ4A ρ
2B . (9.21)
Note that this expression correctly reproduces that last branches of the complete results
(9.11) and (9.20) for 1 and 2 vector multiplets.
For completely generic values of the parameters ρA, it is possible to extract a weaker
information from the constraints (9.2) by noticing that
K(ǫ2A, ρ2A) ≤ 1 + 4 ρ2 ǫ2 + 4 (ρ2 − 1) ǫ4 , (9.22)
with:
ǫ2 =∑
Aǫ2A , ρ2 =
∑
Aρ2A . (9.23)
One can then use the upper bound (9.22) instead of the original expression (9.3) in the
constraints. Doing so, one clearly looses any distinction among the various different mul-
tiplets, and one obtains instead a condition where the effect of the multiplets is somehow
averaged. In fact, the expression (9.22) has the same form as the expression (9.8) that is
32
valid in the case of a single vector multiplet, but with ǫ2 and ρ2 given by eqs. (9.23). The
constrained maximum of (9.22) within the range ǫ2 ≤ ρ2 is located at
ǫ2best =
1
2
ρ2
1 − ρ2, ρ2 <
1
2,
ρ2 , ρ2 >1
2.
(9.24)
In terms of the original variables, this means the optimal situation is the one in which all
the dA are such that the quantity∑
A d2A takes the maximal value
∑
A ρ2A/(1 +
∑
B ρ2B)
that it is allowed by the kinematical bound. As we already mentioned, there is a whole
m-sphere of such directions, and no specific direction is singled out as optimal. This is
due to the fact that in the present analysis any information distinguishing the different
vector multiplets has been disregarded from the beginning. For the function K one finds
Kbest ≤
1 − ρ2 + ρ4
1 − ρ2, ρ2 <
1
2,
1 + 4 ρ6 , ρ2 >1
2.
(9.25)
A last way to derive a weaker constraint that is valid for arbitrary values of the
parameters ρA is to neglect the negative term in K. One is then left with a monotonically
growing function of the variables ǫA, which obviously takes its maximal value at the
boundary ǫA = ρA of the allowed domain. This leads to:
Kbest = 1 + 4∑
Aρ4A + 4
∑
A,Bρ4A ρ
2B . (9.26)
This condition is completely general, and still keeps a distinction between the different
vector multiplets. However, it is generically much weaker that the conditions we got in
the previous analysis.
9.4 Simple scalar geometries
It is worth discussing more precisely the minimization of the term involving the effective
Riemann tensor in the stability condition in the particular cases of separable or symmetric
manifolds where, as already mentioned in section 7, the situation simplifies [17, 18].
For factorizable manifolds, the Riemann tensor can be written in terms of the n curva-
ture scalars RI of the one-dimensional factors and is just given by RIJP Q = RI δIJP Q. The
effective curvature that is relevant for the constraints (9.6) has then the same factorized
form, and is given by RIJP Q = RI δIJP Q, where
RI = Kbest(ρ2A)−1RI . (9.27)
The values of the variables zI optimizing (9.6) are given by |zI |2 = R−1I /
∑
J R−1J . In
terms of the original variables, this means that the f I align along the direction with
highest total effective inverse curvature. One finds then the following necessary condition
on the effective inverse curvature scalars:∑
IR−1I ≥ 3
2. (9.28)
33
For maximally symmetric manifolds, the Riemann tensor can be written in terms of a
single curvature scale Rall as RIJP Q = Rall/2 (δIJδPQ + δIQδP J). The effective curvature
appearing in the flatness and stability conditions (9.6) remains again of the same form,
and reads RIJP Q = Rall/2 (δIJ δPQ + δIQδP J), where:
Rall = Kbest(ρ2A)−1Rall . (9.29)
In this case, no particular direction in the variables zI , or equivalently in the f I , is singled
out as optimal, due to the fact that the space is maximally symmetric and all the directions
are thus equivalent. One finds however the necessary condition:
R−1all ≥ 3
2. (9.30)
10 String inspired examples
In this section we will apply our results to the typical situations arising for the moduli
sector of string models. In order to get a better idea of the usefulness of the conditions
(6.4) for finding phenomenologically viable vacua in the presence of D-terms, we will
study in more detail the form that these conditions take in the simplest case involving 1
chiral and 1 vector multiplet, and in the next-to-simplest case of 2 chiral multiplets with
a factorized geometry and 1 vector multiplet. We will also briefly comment on how the
idea of uplifting introduced in [7] fits into our study.
10.1 One chiral superfield and one isometry
The simplest possible situation arises in a low energy effective theory with one chiral
superfield and one isometry, which is gauged with one vector multiplet. For moduli fields
of string models, the prototype of Kahler potential describing such a situation is of the
form:
K = −n log(
Φ + Φ†)
. (10.1)
This is a constant curvature manifold with R = 2/n. It has a global symmetry associated
to the Killing vector X = i ξ, which can be gauged as long as the superpotential is also
gauge invariant.
In this case, the flatness condition can be solved by introducing an angle δ and
parametrizing the auxiliary fields defined in (6.1) as f = cos δ and d = sin δ. The angle δ
is, in this simple case, fixed by the ratio between the vector and the gravitino masses, the
parameter ρ defined in (6.3):
tan δ = ρ . (10.2)
In this simple situation, the stability condition is actually necessary and sufficient for the
only non-trivial mass eigenvalue to be positive. We can thus change the ≤ sign to a <
sign, and write this condition as
R <2
3
(
1 + 4 ρ6)
. (10.3)
34
From this expression, it is clear that it is always possible to satisfy the stability condition
with a large enough value of ρ. More precisely, recalling that R = 2/n, we get that
n >3
1 + 4 ρ6. (10.4)
Note in particular that eq. (10.4) implies that whenever n is substantially less than 3,
which is the critical value for stability in the absence of gauging, the contribution to
supersymmetry breaking coming from the D auxiliary field must be comparable to the
one coming from the F auxiliary field. This would be for instance the case for an effective
theory based on the dilaton modulus, for which n = 1. On the other hand, if n is close or
equal to 3, like in the case of an effective theory based on the volume modulus, the effect
of D auxiliary fields can be still relevant even if supersymmetry breaking is dominated by
the F auxiliary field.
It is important to note that in this case gauge invariance of G severely restricts the
form of the superpotential, which can only be of the form W = α e−β Φ. With these forms
of K and W , it is easy to see that a satisfactory extremum can actually exist only if n < 3,
as in the absence of the D-term we get a negative definite (n < 3) or positive definite
(n > 3) scalar potential and we need the negative one to compete with the positive definite
contribution due to the D-terms in order to stabilize the field. This model was studied
in detail in [21] 15. Of course, as was already noted in [21], the concept of uplifting
a supersymmetric minimum using D-terms does not apply in this simple case as if we
consider only one chiral field and one vector field the D-term is proportional to the only
available F -term.
10.2 Two chiral superfields and one isometry
Another interesting and reasonably simple case is that of a low energy effective theory
with two chiral superfields with a factorizable geometry and one isometry, which is gauged
with one vector multiplet. For moduli fields in string models, one typically has a Kahler
potential of the form
K = −n1 log (Φ1 + Φ†1) − n2 log (Φ2 + Φ†
2) . (10.6)
The scalar manifold defined by this Kahler potential consists of two one-dimensional sub-
spaces with constant scalar curvatures R1 = 2/n1 and R2 = 2/n2, and has actually two
independent global symmetries, under which the chiral multiplets independently shift by
an imaginary constant. We will consider the case in which we gauge only one linear com-
bination of the isometries, defined by X = (i ξ1, i ξ2). The most general superpotential
allowed by gauge invariance of G is then of the form W = e−β(ξ1Φ1+ξ2Φ2)W (ξ2Φ1 − ξ1Φ2).
15In the more general case of a non-constant gauge kinetic function, which was considered in [21], the
condition (10.3) gets modified. In terms of the parameter κ = g′/(g√
K′′), one finds:
R <2
3
`
1 + 4 ρ6´
+ 4√
3κ ρ4p
1 + ρ2 + 4 κ2 ρ2`
2 + ρ2´
. (10.5)
Again, this condition can always be satisfied for appropriate values of δ.
35
In this case the solution of the flatness condition can be parametrized by two angles
δ and θ, with f1 = cos θ cos δ, f2 = sin θ cos δ and d = sin δ. It is also useful to introduce
the following two angles, defining the orientation of the Killing vector X = (X1,X2) and
the inverse curvature vector R−1 = (R−11 , R−1
2 ):
tanx =ξ2ξ1, tan r =
n2
n1. (10.7)
From the definitions of the vector mass, the gravitino mass, and the parameter ρ, one
finds in this case the following relation between the angles δ and θ:
tan δ
cos (θ − x)= ρ . (10.8)
The stability condition can then be written in the following form:
(
R1 cos4θ +R2 sin4θ)
+2
3
(
ρ4−ρ6)
(
2 cos2(θ−x) − 1
1−ρ2
)2
≤ 2
3
1−ρ2 +ρ4
1−ρ2. (10.9)
In general, this condition cannot be solved exactly, since it is quartic. It is however easy
to see that the first term in the left-hand side of (10.9) takes its minimal value when
cos2θ = R2/(R1 + R2), implying θ = r, whereas the second term gets minimized when
cos2(θ − x) = 1/[2(1 − ρ2)] if ρ ≤ 1/2 and when θ = x if instead ρ > 1/2. One can then
derive a lower bound for the left-hand side by minimizing the two terms separately, and
derive in this way a simple necessary condition. Recalling that R1 = 2/n1 and R2 = 2/n2,
one deduces in this way that:
n1 + n2 ≥ 3
1 − ρ2
1 − ρ2 + ρ4, if ρ ≤ 1/2 ,
1
1 + 4 ρ6, if ρ > 1/2 .
(10.10)
It is clear from this condition that for large enough ρ one can always satisfy the bound, the
minimal value needed for ρ being determined by n1 and n2. Note also that the expression
(10.10) reproduces the same bounds derived in section 9, as is clear from eq. (9.11).
An interesting particular case occurs when one of the shifts vanishes, say ξ1 = 0.
In such a situation, one finds x = π/2, and the condition (10.9) can be solved exactly.
Actually it turns out that for values of ρ such that 3 (n1 + n2) + 4n1n2 ρ4 (1 − ρ2) ≤ 0,
the bound can always be satisfied for some range of values of the angle θ. On the other
hand, for values of ρ such that 3 (n1 + n2) + 4n1n2 ρ4 (1− ρ2) > 0 the following condition
must be satisfied:
n1 + n2 ≥ 3 − 4n2 ρ6 − 4
3n1n2 ρ
4(
1 − ρ2 + ρ4)
. (10.11)
Note also that in such a model, it is possible to stabilize the field Φ1 in a supersymmetric
way. This would correspond to θ = π/2, and the situation would then become identical to
the one described in the previous section. In this kind of model, if the field Φ2 satisfy the
bound (10.3), one could then use the sector containing Φ2 and V to break supersymmetry
and ”uplift” to a Minkowski vacuum a supersymmetric AdS minimum for the sector
involving Φ1.
36
11 Conclusions
In this paper we have studied the constraints that can be put on gauge invariant super-
gravity models from the requirement of the existence of a flat and metastable vacuum.
Following the same strategy as in the previous analysis presented in [17, 18] for the simplest
supergravity theories involving only chiral multiplets, we have considered the constraints
implied by the flatness condition implying the vanishing of the cosmological constant and
by a necessary but not sufficient condition for metastability, derived by looking at the
particular complex direction in the scalar field space that corresponds to the Goldstino
direction. These conditions define two algebraic constraints on the chiral and vector aux-
iliary fields F and D, and depend on the curvature of the Kahler potential, the mass of
the vector fields and the derivatives of the gauge kinetic function. The major difficulty in
solving more explicitly these constraints comes from the kinematical and the dynamical
relations existing between the chiral and vector auxiliary fields. We have presented and
followed three different methods to derive more explicitly the restrictions imposed by the
flatness and stability constraints on the parameters defining the theory. These methods
are based on the use of respectively the dynamical relation, the kinematical relation and
a kinematical bound between the F and D auxiliary fields, and preserve a decreasing
amount of the information contained in the original conditions. In this way, we were able
to obtain several kinds of necessary conditions, which are relevant in different situations.
Our results can be summarized as follows: as expected, the presence of vector multi-
plets, in addition to chiral multiplets, tends to alleviate the constraints with respect to a
situations with only chiral multiplets. This is mainly due to the fact that the D-type aux-
iliary fields give a positive definite contribution to the scalar potential, on the contrary of
the F -type auxiliary fields, which give an indefinite sign contribution. The effect of vector
multiplets is maximal when the gauge boson masses are comparable to the gravitino mass.
When these two mass scales are instead hierarchically different, the effect of vector mul-
tiplets is small and is encoded in a shift of the components of the Riemann tensor. More
in general, we found through various types of analyses that the main effect of the vector
multiplets is essentially to reduce the effective curvature felt by the chiral multiplets, and
thereby to make the condition of metastability less constraining. On the other hand, the
local symmetries associated to the vector multiplets and the corresponding Higgs mecha-
nism controlling their spontaneous breaking imply that the presence of vector fields does
not merely represent an extra generalizing ingredient, but also leads to restrictions.
We believe that the general results derived in this paper can be useful in discriminating
more efficiently potentially viable models among those emerging, for instance, as low-
energy effective descriptions of string models. We leave the exploration of this application
for future work.
Acknowledgments
We thank G. Dall’Agata, E. Dudas, R. Rattazzi, M. Serone and A. Uranga for useful
discussions. This work has been partly supported by the Swiss National Science Founda-
37
tion and by the European Commission under contracts MRTN-CT-2004-005104. We also
thank the Theory Division of CERN for hospitality.
References
[1] E. Witten, Phys. Lett. B 155, 151 (1985);
W. Lang, J. Louis and B. A. Ovrut, Nucl. Phys. B 261, 334 (1985).
[2] J. R. Ellis, C. Kounnas and D. V. Nanopoulos, Nucl. Phys. B 247 (1984) 373;
S. Ferrara, C. Kounnas and M. Porrati, Phys. Lett. B 181 (1986) 263.
[3] J. P. Derendinger, L. E. Ibanez and H. P. Nilles, Phys. Lett. B 155 (1985) 65;
M. Dine, R. Rohm, N. Seiberg and E. Witten, Phys. Lett. B 156 (1985) 55.
[4] S. Gukov, C. Vafa and E. Witten, Nucl. Phys. B 584 (2000) 69 [Errat. B 608 (2001) 477]
[hep-th/9906070].
[5] S. B. Giddings, S. Kachru and J. Polchinski, Phys. Rev. D 66 (2002) 106006 [hep-th/0105097].
[6] L. Susskind, hep-th/0302219;
M. R. Douglas, JHEP 0305 (2003) 046 [hep-th/0303194].
[7] S. Kachru, R. Kallosh, A. Linde and S. P. Trivedi, Phys. Rev. D 68 (2003) 046005 [hep-
th/0301240].
[8] N. Arkani-Hamed, M. Dine and S. P. Martin, Phys. Lett. B 431 (1998) 329 [hep-ph/9803432].
[9] Y. Kawamura, Phys. Lett. B 446 (1999) 228 [hep-ph/9811312].
[10] R. Barbieri, S. Ferrara, D. V. Nanopoulos and K. S. Stelle, Phys. Lett. B 113 (1982) 219;
S. Ferrara, L. Girardello, T. Kugo and A. Van Proeyen, Nucl. Phys. B 223 (1983) 191.
[11] P. Binetruy, G. Dvali, R. Kallosh and A. Van Proeyen, Class. Quant. Grav. 21, 3137 (2004)
[hep-th/0402046].
[12] J. A. Casas, Phys. Lett. B 384 (1996) 103 [hep-th/9605180];
R. Brustein and S. P. de Alwis, Phys. Rev. D 64 (2001) 046004 [hep-th/0002087]; Phys. Rev.
D 69 (2004) 126006 [hep-th/0402088].
[13] T. Barreiro, B. de Carlos and E. J. Copeland, Phys. Rev. D 57 (1998) 7354 [hep-ph/9712443];
C. P. Burgess, J. P. Derendinger, F. Quevedo and M. Quiros, Annals Phys. 250 (1996) 193
[hep-th/9505171].
[14] P. Binetruy, M. K. Gaillard and Y. Y. Wu, Nucl. Phys. B 481 (1996) 109 [hep-th/9605170],
Nucl. Phys. B 493 (1997) 27 [hep-th/9611149].
[15] K. Becker, M. Becker, M. Haack and J. Louis, JHEP 0206 (2002) 060 [hep-th/0204254];
V. Balasubramanian and P. Berglund, JHEP 0411 (2004) 085 [hep-th/0408054];
V. Balasubramanian, P. Berglund, J. P. Conlon and F. Quevedo, JHEP 0503 (2005) 007
[hep-th/0502058];
J. P. Conlon, F. Quevedo and K. Suruliz, JHEP 0508 (2005) 007 [hep-th/0505076].
[16] G. von Gersdorff and A. Hebecker, Phys. Lett. B 624 (2005) 270 [hep-th/0507131];
M. Berg, M. Haack and B. Kors, Phys. Rev. Lett. 96 (2006) 021601 [hep-th/0508171].
[17] M. Gomez-Reino and C. A. Scrucca, JHEP 0605 (2006) 015 [hep-th/0602246].
[18] M. Gomez-Reino and C. A. Scrucca, JHEP 0609 (2006) 008 [hep-th/0606273].
[19] F. Denef and M. R. Douglas, JHEP 0503 (2005) 061 [arXiv:hep-th/0411183].
[20] J. Gray, Y. H. He and A. Lukas, JHEP 0609 (2006) 031 [hep-th/0606122];
J. Gray, Y. H. He, A. Ilderton and A. Lukas, hep-th/0703249.
38
[21] G. Villadoro and F. Zwirner, Phys. Rev. Lett. 95, 231602 (2005) [hep-th/0508167].
[22] G. Villadoro and F. Zwirner, JHEP 0603 (2006) 087 [hep-th/0602120].
[23] A. Achucarro, B. de Carlos, J. A. Casas and L. Doplicher, JHEP 0606 (2006) 014 [hep-
th/0601190].
[24] Z. Lalak, O. J. Eyton-Williams and R. Matyszkiewicz, hep-th/0702026.
[25] E. Dudas and Y. Mambrini, JHEP 0610 (2006) 044 [hep-th/0607077].
[26] S. L. Parameswaran and A. Westphal, JHEP 0610 (2006) 079 [arXiv:hep-th/0602253].
[27] D. Cremades, M. P. Garcia del Moral, F. Quevedo and K. Suruliz, arXiv:hep-th/0701154.
[28] M. A. Luty and R. Sundrum, Phys. Rev. D 62 (2000) 035008 [hep-th/9910202].
[29] R. Rattazzi, C. A. Scrucca and A. Strumia, Nucl. Phys. B 674 (2003) 171 [hep-th/0305184];
T. Gregoire, R. Rattazzi, C. A. Scrucca, A. Strumia and E. Trincherini, Nucl. Phys. B 720
(2005) 3 [hep-th/0411216];
F. Brummer, A. Hebecker and M. Trapletti, Nucl. Phys. B 755 (2006) 186 [hep-th/0605232].
[30] O. Lebedev, H. P. Nilles and M. Ratz, Phys. Lett. B 636 (2006) 126 [hep-th/0603047];
O. Lebedev, V. Lowen, Y. Mambrini, H. P. Nilles and M. Ratz, JHEP 0702 (2007) 063
[hep-ph/0612035];
[31] R. Kallosh and A. Linde, JHEP 0702 (2007) 002 [hep-th/0611183].
[32] E. Dudas, C. Papineau and S. Pokorski, JHEP 0702 (2007) 028 [hep-th/0610297];
H. Abe, T. Higaki, T. Kobayashi and Y. Omura, Phys. Rev. D 75 (2007) 025019 [hep-
th/0611024].
[33] K. Intriligator, N. Seiberg and D. Shih, JHEP 0604 (2006) 021 [hep-th/0602239].
[34] C. P. Burgess, R. Kallosh and F. Quevedo, JHEP 0310 (2003) 056 [hep-th/0309187].
[35] K. Choi, A. Falkowski, H. P. Nilles and M. Olechowski, Nucl. Phys. B 718 (2005) 113 [hep-
th/0503216].
[36] S. P. de Alwis, Phys. Lett. B 626 (2005) 223 [hep-th/0506266].
[37] E. Cremmer et al., Nucl. Phys. B 147 (1979) 105;
E. Witten and J. Bagger, Phys. Lett. B 115 (1982) 202.
[38] E. Cremmer, S. Ferrara, L. Girardello and A. Van Proeyen, Phys. Lett. B 116 (1982) 231;
Nucl. Phys. B 212 (1983) 413;
J. A. Bagger, Nucl. Phys. B 211 (1983) 302.
[39] J. Wess and J. Bagger, Princeton, USA: Univ. Pr. (1992) 259 p
[40] L. Andrianopoli, S. Ferrara and M. A. Lledo, JHEP 0404 (2004) 005 [hep-th/0402142];
J. De Rydt, J. Rosseel, T. T. Schmidt, A. Van Proeyen and M. Zagermann, arXiv:0705.4216;
P. Anastasopoulos, M. Bianchi, E. Dudas and E. Kiritsis, JHEP 0611 (2006) 057 [hep-
th/0605225].
[41] S. Aoyama and J. W. van Holten, Nucl. Phys. B 258 (1985) 18;
W. Buchmuller and W. Lerche, Annals Phys. 175 (1987) 159.
[42] H. Elvang, D. Z. Freedman and B. Kors, JHEP 0611 (2006) 068 [hep-th/0606012];
D. Z. Freedman and B. Kors, JHEP 0611 (2006) 067 [hep-th/0509217].
[43] M. B. Green and J. H. Schwarz, Phys. Lett. B 149 (1984) 117.
[44] M. Kaku, P. K. Townsend and P. van Nieuwenhuizen, Phys. Rev. D 17 (1978) 3179;
S. Ferrara, L. Girardello, T. Kugo and A. Van Proeyen, Nucl. Phys. B 223 (1983) 191;
T. Kugo and S. Uehara, Nucl. Phys. B 226 (1983) 49.
39
[45] S. Ferrara, C. Kounnas and F. Zwirner, Nucl. Phys. B 429 (1994) 589 [Errat. B 433 (1995)
255] [hep-th/9405188].
[46] E. Dudas and S. K. Vempati, Nucl. Phys. B 727 (2005) 139 [hep-th/0506172].
[47] K. Choi and K. S. Jeong, JHEP 0608 (2006) 007 [hep-th/0605108].
[48] Y. Kawamura and T. Kobayashi, Phys. Rev. D 56 (1997) 3844 [hep-ph/9608233]; Phys. Lett.
B 375 (1996) 141 [Erratum-ibid. B 388 (1996) 867] [hep-ph/9601365].
[49] T. Gregoire, R. Rattazzi and C. A. Scrucca, Phys. Lett. B 624 (2005) 260 [hep-ph/0505126].
40